Merge branch 'main' into nmap/components-impl

# Conflicts:
#	management/server/posture/checks.go

client/internal/connect.go

@@ -11,6 +11,7 @@ import (
 	"runtime/debug"
 	"strings"
 	"sync"
+	"sync/atomic"
 	"time"
 
 	"github.com/cenkalti/backoff/v4"
@@ -54,6 +55,10 @@ var androidRunOverride func(c *ConnectClient, runningChan chan struct{}, logPath
 
 type ConnectClient struct {
 	ctx            context.Context
+	runCancel      context.CancelFunc
+	runExited      chan struct{}
+	runOnce        sync.Once
+	runStarted     atomic.Bool
 	config         *profilemanager.Config
 	statusRecorder *peer.Status
 
@@ -70,8 +75,14 @@ func NewConnectClient(
 	config *profilemanager.Config,
 	statusRecorder *peer.Status,
 ) *ConnectClient {
+	// Derive the run context here so Stop owns the cancel that unblocks the run
+	// loop. runCancel is set once at construction, so Stop can call it without
+	// racing the run loop's startup. Callers therefore need not cancel before Stop.
+	runCtx, runCancel := context.WithCancel(ctx)
 	return &ConnectClient{
-		ctx:            ctx,
+		ctx:            runCtx,
+		runCancel:      runCancel,
+		runExited:      make(chan struct{}),
 		config:         config,
 		statusRecorder: statusRecorder,
 		engineMutex:    sync.Mutex{},
@@ -135,6 +146,11 @@ func (c *ConnectClient) RunOniOS(
 }
 
 func (c *ConnectClient) run(mobileDependency MobileDependency, runningChan chan struct{}, logPath string) error {
+	// Mark the loop as started and signal exit on return so Stop can wait for
+	// the loop to finish (and skip the wait if the loop never ran).
+	c.runStarted.Store(true)
+	defer c.runOnce.Do(func() { close(c.runExited) })
+
 	defer func() {
 		if r := recover(); r != nil {
 			rec := c.statusRecorder
@@ -290,7 +306,7 @@ func (c *ConnectClient) run(mobileDependency MobileDependency, runningChan chan
 			log.Debug(err)
 			if s, ok := gstatus.FromError(err); ok && (s.Code() == codes.PermissionDenied) {
 				state.Set(StatusNeedsLogin)
-				_ = c.Stop()
+				c.runCancel()
 				return backoff.Permanent(wrapErr(err)) // unrecoverable error
 			}
 			return wrapErr(err)
@@ -410,14 +426,10 @@ func (c *ConnectClient) run(mobileDependency MobileDependency, runningChan chan
 		c.engine = nil
 		c.engineMutex.Unlock()
 
-		// todo: consider to remove this condition. Is not thread safe.
-		// We should always call Stop(), but we need to verify that it is idempotent
-		if engine.wgInterface != nil {
-			log.Infof("ensuring %s is removed, Netbird engine context cancelled", engine.wgInterface.Name())
+		log.Infof("ensuring wg interface is removed, Netbird engine context cancelled")
 
-			if err := engine.Stop(); err != nil {
-				log.Errorf("Failed to stop engine: %v", err)
-			}
+		if err := engine.Stop(); err != nil {
+			log.Errorf("Failed to stop engine: %v", err)
 		}
 		c.statusRecorder.ClientTeardown()
 
@@ -433,12 +445,12 @@ func (c *ConnectClient) run(mobileDependency MobileDependency, runningChan chan
 	}
 
 	c.statusRecorder.ClientStart()
-	err = backoff.Retry(operation, backOff)
+	err = backoff.Retry(operation, backoff.WithContext(backOff, c.ctx))
 	if err != nil {
 		log.Debugf("exiting client retry loop due to unrecoverable error: %s", err)
 		if s, ok := gstatus.FromError(err); ok && (s.Code() == codes.PermissionDenied) {
 			state.Set(StatusNeedsLogin)
-			_ = c.Stop()
+			c.runCancel()
 		}
 		return err
 	}
@@ -516,11 +528,9 @@ func (c *ConnectClient) Status() StatusType {
 }
 
 func (c *ConnectClient) Stop() error {
-	engine := c.Engine()
-	if engine != nil {
-		if err := engine.Stop(); err != nil {
-			return fmt.Errorf("stop engine: %w", err)
-		}
+	c.runCancel()
+	if c.runStarted.Load() {
+		<-c.runExited
 	}
 	return nil
 }

client/internal/debug/debug_test.go

@@ -843,6 +843,7 @@ func TestAddConfig_AllFieldsCovered(t *testing.T) {
 		"PreSharedKey":      "sensitive: WireGuard pre-shared key",
 		"SSHKey":            "sensitive: SSH private key",
 		"ClientCertKeyPair": "non-config: parsed cert pair, not serialized",
+		"Name":              "non-config: profile name is not needed for debug purposes",
 		"policy":            "non-config: in-memory MDM policy snapshot, surfaced via Config.Policy() / GetConfigResponse.MDMManagedFields",
 	}
 

client/internal/dns/mgmt/mgmt.go

@@ -51,13 +51,20 @@ type cachedRecord struct {
 }
 
 // Resolver caches critical NetBird infrastructure domains.
-// records, refreshing, mgmtDomain and serverDomains are all guarded by mutex.
+// records, refreshing, failedResolves, mgmtDomain and serverDomains are all
+// guarded by mutex.
 type Resolver struct {
 	records       map[dns.Question]*cachedRecord
 	mgmtDomain    *domain.Domain
 	serverDomains *dnsconfig.ServerDomains
 	mutex         sync.RWMutex
 
+	// failedResolves records the last failed initial resolve per domain so a
+	// domain that never resolves isn't retried on every server-domains update
+	// until refreshBackoff elapses. Entries are cleared on success and pruned
+	// to the current server-domains set.
+	failedResolves map[domain.Domain]time.Time
+
 	chain            ChainResolver
 	chainMaxPriority int
 	refreshGroup     singleflight.Group
@@ -76,9 +83,10 @@ type Resolver struct {
 // NewResolver creates a new management domains cache resolver.
 func NewResolver() *Resolver {
 	return &Resolver{
-		records:    make(map[dns.Question]*cachedRecord),
-		refreshing: make(map[dns.Question]*atomic.Bool),
-		cacheTTL:   resolveCacheTTL(),
+		records:        make(map[dns.Question]*cachedRecord),
+		refreshing:     make(map[dns.Question]*atomic.Bool),
+		failedResolves: make(map[domain.Domain]time.Time),
+		cacheTTL:       resolveCacheTTL(),
 	}
 }
 
@@ -173,7 +181,9 @@ func (m *Resolver) continueToNext(w dns.ResponseWriter, r *dns.Msg) {
 
 // AddDomain resolves a domain and stores its A/AAAA records in the cache.
 // A family that resolves NODATA (nil err, zero records) evicts any stale
-// entry for that qtype.
+// entry for that qtype. When one family hard-errors while the other succeeds,
+// the resolved family is still cached but AddDomain returns an error so the
+// caller retries the incomplete resolve rather than treating it as complete.
 func (m *Resolver) AddDomain(ctx context.Context, d domain.Domain) error {
 	dnsName := strings.ToLower(dns.Fqdn(d.PunycodeString()))
 
@@ -203,6 +213,10 @@ func (m *Resolver) AddDomain(ctx context.Context, d domain.Domain) error {
 	log.Debugf("added/updated domain=%s with %d A records and %d AAAA records",
 		d.SafeString(), len(aRecords), len(aaaaRecords))
 
+	if errA != nil || errAAAA != nil {
+		return fmt.Errorf("resolve %s: incomplete, a family failed: %w", d.SafeString(), errors.Join(errA, errAAAA))
+	}
+
 	return nil
 }
 
@@ -462,6 +476,7 @@ func (m *Resolver) RemoveDomain(d domain.Domain) error {
 	delete(m.records, qAAAA)
 	delete(m.refreshing, qA)
 	delete(m.refreshing, qAAAA)
+	delete(m.failedResolves, d)
 
 	log.Debugf("removed domain=%s from cache", d.SafeString())
 	return nil
@@ -505,6 +520,7 @@ func (m *Resolver) UpdateFromServerDomains(ctx context.Context, serverDomains dn
 		allDomains := m.extractDomainsFromServerDomains(updatedServerDomains)
 		currentDomains := m.GetCachedDomains()
 		removedDomains = m.removeStaleDomains(currentDomains, allDomains)
+		m.pruneFailedResolves(allDomains)
 	}
 
 	m.addNewDomains(ctx, newDomains)
@@ -577,13 +593,85 @@ func (m *Resolver) isManagementDomain(domain domain.Domain) bool {
 	return m.mgmtDomain != nil && domain == *m.mgmtDomain
 }
 
-// addNewDomains resolves and caches all domains from the update
+// addNewDomains resolves and caches domains that are not yet in the cache,
+// running the lookups concurrently. Domains already cached are skipped and left
+// to the stale-while-revalidate refresh path, so a sync never re-resolves them
+// synchronously: once NetBird owns the OS resolver the resolve runs through the
+// handler chain and would otherwise dial the managed upstreams under the engine
+// sync lock on every update.
 func (m *Resolver) addNewDomains(ctx context.Context, newDomains domain.List) {
+	var wg sync.WaitGroup
+	seen := make(map[domain.Domain]struct{}, len(newDomains))
 	for _, newDomain := range newDomains {
-		if err := m.AddDomain(ctx, newDomain); err != nil {
-			log.Warnf("failed to add/update domain=%s: %v", newDomain.SafeString(), err)
-		} else {
-			log.Debugf("added/updated management cache domain=%s", newDomain.SafeString())
+		if _, dup := seen[newDomain]; dup {
+			continue
+		}
+		seen[newDomain] = struct{}{}
+
+		if !m.needsResolve(newDomain) {
+			continue
+		}
+
+		wg.Add(1)
+		go func(d domain.Domain) {
+			defer wg.Done()
+			if err := m.AddDomain(ctx, d); err != nil {
+				m.markResolveFailed(d)
+				log.Warnf("failed to add/update domain=%s: %v", d.SafeString(), err)
+				return
+			}
+			m.clearResolveFailed(d)
+			log.Debugf("added/updated management cache domain=%s", d.SafeString())
+		}(newDomain)
+	}
+	wg.Wait()
+}
+
+// needsResolve reports whether d should be resolved now. A recent failed or
+// incomplete resolve gates retries on the backoff even when one family is
+// already cached, so a transiently-failed family is retried instead of being
+// treated as fully resolved. Otherwise a domain with any cached record is left
+// to the stale-while-revalidate refresh path.
+func (m *Resolver) needsResolve(d domain.Domain) bool {
+	dnsName := strings.ToLower(dns.Fqdn(d.PunycodeString()))
+
+	m.mutex.RLock()
+	defer m.mutex.RUnlock()
+
+	if failedAt, ok := m.failedResolves[d]; ok {
+		return time.Since(failedAt) >= refreshBackoff
+	}
+
+	for _, qtype := range []uint16{dns.TypeA, dns.TypeAAAA} {
+		q := dns.Question{Name: dnsName, Qtype: qtype, Qclass: dns.ClassINET}
+		if _, ok := m.records[q]; ok {
+			return false
+		}
+	}
+	return true
+}
+
+func (m *Resolver) markResolveFailed(d domain.Domain) {
+	m.mutex.Lock()
+	m.failedResolves[d] = time.Now()
+	m.mutex.Unlock()
+}
+
+func (m *Resolver) clearResolveFailed(d domain.Domain) {
+	m.mutex.Lock()
+	delete(m.failedResolves, d)
+	m.mutex.Unlock()
+}
+
+// pruneFailedResolves drops failure markers for domains no longer present in
+// the server-domains set, keeping the map bounded to the current set (a
+// failed-only domain has no cached record, so RemoveDomain never sees it).
+func (m *Resolver) pruneFailedResolves(domains domain.List) {
+	m.mutex.Lock()
+	defer m.mutex.Unlock()
+	for d := range m.failedResolves {
+		if !slices.Contains(domains, d) {
+			delete(m.failedResolves, d)
 		}
 	}
 }

client/internal/dns/mgmt/mgmt_refresh_test.go

@@ -21,6 +21,7 @@ type fakeChain struct {
 	mu       sync.Mutex
 	calls    map[string]int
 	answers  map[string][]dns.RR
+	qErr     map[string]error
 	err      error
 	hasRoot  bool
 	onLookup func()
@@ -30,6 +31,7 @@ func newFakeChain() *fakeChain {
 	return &fakeChain{
 		calls:   map[string]int{},
 		answers: map[string][]dns.RR{},
+		qErr:    map[string]error{},
 		hasRoot: true,
 	}
 }
@@ -47,6 +49,9 @@ func (f *fakeChain) ResolveInternal(ctx context.Context, msg *dns.Msg, maxPriori
 	f.calls[key]++
 	answers := f.answers[key]
 	err := f.err
+	if err == nil {
+		err = f.qErr[key]
+	}
 	onLookup := f.onLookup
 	f.mu.Unlock()
 
@@ -75,6 +80,12 @@ func (f *fakeChain) setAnswer(name string, qtype uint16, ip string) {
 	}
 }
 
+func (f *fakeChain) setErr(name string, qtype uint16, err error) {
+	f.mu.Lock()
+	defer f.mu.Unlock()
+	f.qErr[name+"|"+dns.TypeToString[qtype]] = err
+}
+
 func (f *fakeChain) callCount(name string, qtype uint16) int {
 	f.mu.Lock()
 	defer f.mu.Unlock()

client/internal/dns/mgmt/mgmt_resolve_test.go

@@ -0,0 +1,183 @@
+package mgmt
+
+import (
+	"context"
+	"errors"
+	"sync/atomic"
+	"testing"
+	"time"
+
+	"github.com/miekg/dns"
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+
+	dnsconfig "github.com/netbirdio/netbird/client/internal/dns/config"
+	"github.com/netbirdio/netbird/shared/management/domain"
+)
+
+// A domain already in the cache must not be re-resolved on a subsequent server
+// domains update; it is left to the stale-while-revalidate refresh path.
+func TestResolver_UpdateFromServerDomains_SkipsCached(t *testing.T) {
+	r := NewResolver()
+	chain := newFakeChain()
+	chain.setAnswer("signal.example.com.", dns.TypeA, "10.0.0.2")
+	r.SetChainResolver(chain, 50)
+
+	sd := dnsconfig.ServerDomains{Signal: domain.Domain("signal.example.com")}
+
+	_, err := r.UpdateFromServerDomains(context.Background(), sd)
+	require.NoError(t, err)
+	require.Equal(t, 1, chain.callCount("signal.example.com.", dns.TypeA),
+		"first update must resolve the domain")
+
+	_, err = r.UpdateFromServerDomains(context.Background(), sd)
+	require.NoError(t, err)
+	assert.Equal(t, 1, chain.callCount("signal.example.com.", dns.TypeA),
+		"cached domain must not be re-resolved on a subsequent update")
+}
+
+// New domains in a single update must resolve concurrently rather than serially.
+func TestResolver_AddNewDomains_ResolvesConcurrently(t *testing.T) {
+	r := NewResolver()
+	chain := newFakeChain()
+
+	var inflight, maxInflight atomic.Int32
+	chain.onLookup = func() {
+		n := inflight.Add(1)
+		for {
+			old := maxInflight.Load()
+			if n <= old || maxInflight.CompareAndSwap(old, n) {
+				break
+			}
+		}
+		time.Sleep(50 * time.Millisecond)
+		inflight.Add(-1)
+	}
+
+	relays := []domain.Domain{"a.example.com", "b.example.com", "c.example.com", "d.example.com"}
+	for _, d := range relays {
+		chain.setAnswer(dns.Fqdn(string(d)), dns.TypeA, "10.0.0.2")
+	}
+	r.SetChainResolver(chain, 50)
+
+	start := time.Now()
+	_, err := r.UpdateFromServerDomains(context.Background(), dnsconfig.ServerDomains{Relay: relays})
+	require.NoError(t, err)
+	elapsed := time.Since(start)
+
+	assert.GreaterOrEqual(t, int(maxInflight.Load()), 2, "domains must resolve concurrently")
+	// Serial resolution of 4 domains would take at least 4*50ms; concurrent is far less.
+	assert.Less(t, elapsed, 300*time.Millisecond, "resolution should not be serial")
+}
+
+// A domain that fails to resolve must not be retried on every update; the
+// failure backoff suppresses re-resolution until it expires.
+func TestResolver_UpdateFromServerDomains_BacksOffFailures(t *testing.T) {
+	r := NewResolver()
+	chain := newFakeChain()
+	chain.err = errors.New("resolve boom")
+	r.SetChainResolver(chain, 50)
+
+	sd := dnsconfig.ServerDomains{Signal: domain.Domain("signal.example.com")}
+
+	_, err := r.UpdateFromServerDomains(context.Background(), sd)
+	require.NoError(t, err)
+	require.Equal(t, 1, chain.callCount("signal.example.com.", dns.TypeA),
+		"first update must attempt the resolve")
+
+	_, err = r.UpdateFromServerDomains(context.Background(), sd)
+	require.NoError(t, err)
+	assert.Equal(t, 1, chain.callCount("signal.example.com.", dns.TypeA),
+		"failed resolve must back off and not retry on the next update")
+}
+
+// A domain listed under more than one server-domain type (e.g. STUN and TURN on
+// the same host) must be resolved once per update, not once per occurrence.
+func TestResolver_AddNewDomains_DedupesDuplicateDomains(t *testing.T) {
+	r := NewResolver()
+	chain := newFakeChain()
+	chain.setAnswer("dup.example.com.", dns.TypeA, "10.0.0.9")
+	r.SetChainResolver(chain, 50)
+
+	sd := dnsconfig.ServerDomains{
+		Stuns: []domain.Domain{"dup.example.com"},
+		Turns: []domain.Domain{"dup.example.com"},
+	}
+
+	_, err := r.UpdateFromServerDomains(context.Background(), sd)
+	require.NoError(t, err)
+	assert.Equal(t, 1, chain.callCount("dup.example.com.", dns.TypeA),
+		"a domain appearing under multiple server-domain types must resolve once")
+}
+
+// A failure marker must be dropped once its domain leaves the server-domains set
+// so the map stays bounded to the current set.
+func TestResolver_UpdateFromServerDomains_PrunesFailedResolves(t *testing.T) {
+	r := NewResolver()
+	chain := newFakeChain()
+	chain.err = errors.New("resolve boom")
+	r.SetChainResolver(chain, 50)
+
+	_, err := r.UpdateFromServerDomains(context.Background(), dnsconfig.ServerDomains{Signal: domain.Domain("gone.example.com")})
+	require.NoError(t, err)
+	r.mutex.RLock()
+	_, marked := r.failedResolves[domain.Domain("gone.example.com")]
+	r.mutex.RUnlock()
+	require.True(t, marked, "failed resolve must be recorded")
+
+	_, err = r.UpdateFromServerDomains(context.Background(), dnsconfig.ServerDomains{Signal: domain.Domain("other.example.com")})
+	require.NoError(t, err)
+	r.mutex.RLock()
+	_, stillMarked := r.failedResolves[domain.Domain("gone.example.com")]
+	r.mutex.RUnlock()
+	assert.False(t, stillMarked, "failure marker for a domain no longer in the set must be pruned")
+}
+
+// When one family hard-errors while the other resolves, the domain is cached
+// for the working family but recorded as incomplete so the failed family is
+// retried under backoff instead of being treated as fully resolved forever.
+func TestResolver_AddNewDomains_RetriesPartialFamilyFailure(t *testing.T) {
+	d := domain.Domain("relay.example.com")
+	r := NewResolver()
+	chain := newFakeChain()
+	chain.setAnswer("relay.example.com.", dns.TypeA, "10.0.0.2")
+	chain.setErr("relay.example.com.", dns.TypeAAAA, errors.New("servfail"))
+	r.SetChainResolver(chain, 50)
+
+	_, err := r.UpdateFromServerDomains(context.Background(), dnsconfig.ServerDomains{Relay: []domain.Domain{d}})
+	require.NoError(t, err)
+
+	r.mutex.RLock()
+	_, aCached := r.records[dns.Question{Name: "relay.example.com.", Qtype: dns.TypeA, Qclass: dns.ClassINET}]
+	_, marked := r.failedResolves[d]
+	r.mutex.RUnlock()
+	require.True(t, aCached, "the working family must still be cached")
+	require.True(t, marked, "a partial failure must be recorded so the failed family is retried")
+
+	assert.False(t, r.needsResolve(d), "within the backoff window the domain is not retried")
+
+	r.mutex.Lock()
+	r.failedResolves[d] = time.Now().Add(-2 * refreshBackoff)
+	r.mutex.Unlock()
+	assert.True(t, r.needsResolve(d), "after the backoff elapses the domain is retried to pick up the missing family")
+}
+
+// A family that returns NODATA (legitimately absent, e.g. an IPv4-only host) is
+// not a failure: the domain must not be marked for retry, otherwise it would be
+// re-resolved on every sync.
+func TestResolver_AddNewDomains_NodataIsNotFailure(t *testing.T) {
+	d := domain.Domain("v4only.example.com")
+	r := NewResolver()
+	chain := newFakeChain()
+	chain.setAnswer("v4only.example.com.", dns.TypeA, "10.0.0.2")
+	r.SetChainResolver(chain, 50)
+
+	_, err := r.UpdateFromServerDomains(context.Background(), dnsconfig.ServerDomains{Relay: []domain.Domain{d}})
+	require.NoError(t, err)
+
+	r.mutex.RLock()
+	_, marked := r.failedResolves[d]
+	r.mutex.RUnlock()
+	assert.False(t, marked, "a NODATA family must not be recorded as a failure")
+	assert.False(t, r.needsResolve(d), "an IPv4-only host must not be re-resolved on later syncs")
+}

client/internal/dns/resutil/resolve.go

@@ -207,3 +207,35 @@ func FormatAnswers(answers []dns.RR) string {
 	}
 	return "[" + strings.Join(parts, ", ") + "]"
 }
+
+// StripOPT removes any OPT pseudo-RRs from the message's Extra section. Per
+// RFC 6891 a responder must not include an OPT RR toward a client that did not
+// advertise EDNS0.
+func StripOPT(msg *dns.Msg) {
+	if len(msg.Extra) == 0 {
+		return
+	}
+	out := msg.Extra[:0]
+	for _, rr := range msg.Extra {
+		if _, ok := rr.(*dns.OPT); ok {
+			continue
+		}
+		out = append(out, rr)
+	}
+	msg.Extra = out
+}
+
+// ExtractEDE returns the first Extended DNS Error (RFC 8914) option carried in
+// the message, if present.
+func ExtractEDE(msg *dns.Msg) (*dns.EDNS0_EDE, bool) {
+	opt := msg.IsEdns0()
+	if opt == nil {
+		return nil, false
+	}
+	for _, o := range opt.Option {
+		if ede, ok := o.(*dns.EDNS0_EDE); ok {
+			return ede, true
+		}
+	}
+	return nil, false
+}

client/internal/dns/resutil/resolve_test.go

@@ -120,3 +120,42 @@ func TestLookupIP_DNSErrorNotIsNotFound(t *testing.T) {
 
 	assert.Equal(t, dns.RcodeServerFailure, result.Rcode, "upstream failure should map to SERVFAIL")
 }
+
+func TestStripOPT(t *testing.T) {
+	rm := &dns.Msg{
+		Extra: []dns.RR{
+			&dns.OPT{Hdr: dns.RR_Header{Name: ".", Rrtype: dns.TypeOPT}},
+			&dns.A{Hdr: dns.RR_Header{Name: "x.", Rrtype: dns.TypeA}, A: net.IPv4(1, 2, 3, 4)},
+		},
+	}
+	StripOPT(rm)
+	assert.Len(t, rm.Extra, 1, "OPT should be removed, A kept")
+	_, isOPT := rm.Extra[0].(*dns.OPT)
+	assert.False(t, isOPT, "remaining record must not be OPT")
+}
+
+func TestExtractEDE(t *testing.T) {
+	t.Run("no edns", func(t *testing.T) {
+		_, ok := ExtractEDE(&dns.Msg{})
+		assert.False(t, ok, "message without OPT has no EDE")
+	})
+
+	t.Run("edns without ede", func(t *testing.T) {
+		rm := &dns.Msg{}
+		rm.SetEdns0(4096, false)
+		_, ok := ExtractEDE(rm)
+		assert.False(t, ok, "OPT without EDE option returns false")
+	})
+
+	t.Run("with ede", func(t *testing.T) {
+		rm := &dns.Msg{}
+		opt := &dns.OPT{Hdr: dns.RR_Header{Name: ".", Rrtype: dns.TypeOPT}}
+		opt.Option = append(opt.Option, &dns.EDNS0_EDE{InfoCode: 49152, ExtraText: "upstream timeout"})
+		rm.Extra = append(rm.Extra, opt)
+
+		ede, ok := ExtractEDE(rm)
+		assert.True(t, ok, "EDE option should be found")
+		assert.Equal(t, uint16(49152), ede.InfoCode)
+		assert.Equal(t, "upstream timeout", ede.ExtraText)
+	})
+}

client/internal/dns/server.go

@@ -6,6 +6,7 @@ import (
 	"fmt"
 	"net/netip"
 	"net/url"
+	"os"
 	"slices"
 	"strings"
 	"sync"
@@ -38,11 +39,15 @@ const (
 	// defaultWarningDelayBase is the starting grace window before a
 	// "Nameserver group unreachable" event fires for a group that's
 	// never been healthy and only has overlay upstreams with no
-	// Connected peer. Per-server and overridable; see warningDelayFor.
-	defaultWarningDelayBase = 30 * time.Second
+	// Connected peer. Per-server and overridable via envWarningDelay;
+	// see warningDelay.
+	defaultWarningDelayBase = 60 * time.Second
 	// warningDelayBonusCap caps the route-count bonus added to the
-	// base grace window. See warningDelayFor.
+	// base grace window. See warningDelay.
 	warningDelayBonusCap = 30 * time.Second
+	// envWarningDelay overrides defaultWarningDelayBase with a Go duration
+	// string (e.g. "90s", "2m"). Invalid or non-positive values are ignored.
+	envWarningDelay = "NB_DNS_HEALTH_WARNING_DELAY"
 )
 
 // errNoUsableNameservers signals that a merged-domain group has no usable
@@ -135,7 +140,7 @@ type DefaultServer struct {
 	disableSys         bool
 	mux                sync.Mutex
 	service            service
-	dnsMuxMap          registeredHandlerMap
+	dnsMuxHandlers     []handlerWrapper
 	localResolver      *local.Resolver
 	wgInterface        WGIface
 	hostManager        hostManager
@@ -199,8 +204,6 @@ type handlerWrapper struct {
 	priority int
 }
 
-type registeredHandlerMap map[types.HandlerID]handlerWrapper
-
 // DefaultServerConfig holds configuration parameters for NewDefaultServer
 type DefaultServerConfig struct {
 	WgInterface    WGIface
@@ -289,7 +292,6 @@ func newDefaultServer(
 		service:           dnsService,
 		handlerChain:      handlerChain,
 		extraDomains:      make(map[domain.Domain]int),
-		dnsMuxMap:         make(registeredHandlerMap),
 		localResolver:     local.NewResolver(),
 		wgInterface:       wgInterface,
 		statusRecorder:    statusRecorder,
@@ -298,7 +300,7 @@ func newDefaultServer(
 		hostManager:       &noopHostConfigurator{},
 		mgmtCacheResolver: mgmtCacheResolver,
 		currentConfigHash: ^uint64(0), // Initialize to max uint64 to ensure first config is always applied
-		warningDelayBase:  defaultWarningDelayBase,
+		warningDelayBase:  warningDelayBaseFromEnv(),
 		healthRefresh:     make(chan struct{}, 1),
 	}
 	// Wire the local resolver against the peer status recorder so it can
@@ -328,7 +330,7 @@ func (s *DefaultServer) SetRouteSources(selected, active func() route.HAMap) {
 	type routeSettable interface {
 		setSelectedRoutes(func() route.HAMap)
 	}
-	for _, entry := range s.dnsMuxMap {
+	for _, entry := range s.dnsMuxHandlers {
 		if h, ok := entry.handler.(routeSettable); ok {
 			h.setSelectedRoutes(selected)
 		}
@@ -978,19 +980,23 @@ func (s *DefaultServer) usableNameServers(nameServers []nbdns.NameServer) []neti
 
 func (s *DefaultServer) updateMux(muxUpdates []handlerWrapper) {
 	// this will introduce a short period of time when the server is not able to handle DNS requests
-	for _, existing := range s.dnsMuxMap {
+	for _, existing := range s.dnsMuxHandlers {
 		s.deregisterHandler([]string{existing.domain}, existing.priority)
-		existing.handler.Stop()
+		// The local resolver is a persistent singleton shared by every custom
+		// zone and reused across config updates. Its chain registrations are
+		// per-config and must be deregistered, but Stop() cancels its lookup
+		// context (breaking external CNAME-target resolution) and clears its
+		// records, so it must not be torn down here.
+		if existing.handler != s.localResolver {
+			existing.handler.Stop()
+		}
 	}
 
-	muxUpdateMap := make(registeredHandlerMap)
-
 	for _, update := range muxUpdates {
 		s.registerHandler([]string{update.domain}, update.handler, update.priority)
-		muxUpdateMap[update.handler.ID()] = update
 	}
 
-	s.dnsMuxMap = muxUpdateMap
+	s.dnsMuxHandlers = muxUpdates
 }
 
 // updateNSGroupStates records the new group set and pokes the refresher.
@@ -1154,6 +1160,26 @@ func (s *DefaultServer) projectUnhealthy(p *nsGroupProj, servers []netip.AddrPor
 	return false
 }
 
+// warningDelayBaseFromEnv returns the base grace window, honoring
+// envWarningDelay when it holds a valid positive Go duration. Invalid or
+// non-positive values fall back to defaultWarningDelayBase.
+func warningDelayBaseFromEnv() time.Duration {
+	val := os.Getenv(envWarningDelay)
+	if val == "" {
+		return defaultWarningDelayBase
+	}
+	d, err := time.ParseDuration(val)
+	if err != nil {
+		log.Warnf("invalid %s value %q, using default %v: %v", envWarningDelay, val, defaultWarningDelayBase, err)
+		return defaultWarningDelayBase
+	}
+	if d <= 0 {
+		log.Warnf("%s must be positive, got %v, using default %v", envWarningDelay, d, defaultWarningDelayBase)
+		return defaultWarningDelayBase
+	}
+	return d
+}
+
 // warningDelay returns the grace window for the given selected-route
 // count. Scales gently: +1s per 100 routes, capped by
 // warningDelayBonusCap. Parallel handshakes mean handshake time grows
@@ -1204,7 +1230,7 @@ func (s *DefaultServer) groupHasImmediateUpstream(servers []netip.AddrPort, snap
 // in more than one handler.
 func (s *DefaultServer) collectUpstreamHealth() map[netip.AddrPort]UpstreamHealth {
 	merged := make(map[netip.AddrPort]UpstreamHealth)
-	for _, entry := range s.dnsMuxMap {
+	for _, entry := range s.dnsMuxHandlers {
 		reporter, ok := entry.handler.(upstreamHealthReporter)
 		if !ok {
 			continue

client/internal/dns/server_test.go

@@ -104,19 +104,6 @@ func init() {
 	formatter.SetTextFormatter(log.StandardLogger())
 }
 
-func generateDummyHandler(d string, servers []nbdns.NameServer) *upstreamResolverBase {
-	var srvs []netip.AddrPort
-	for _, srv := range servers {
-		srvs = append(srvs, srv.AddrPort())
-	}
-	u := &upstreamResolverBase{
-		domain: domain.Domain(d),
-		cancel: func() {},
-	}
-	u.addRace(srvs)
-	return u
-}
-
 func TestUpdateDNSServer(t *testing.T) {
 
 	nameServers := []nbdns.NameServer{
@@ -132,22 +119,20 @@ func TestUpdateDNSServer(t *testing.T) {
 		},
 	}
 
-	dummyHandler := local.NewResolver()
-
 	testCases := []struct {
 		name                string
-		initUpstreamMap     registeredHandlerMap
+		initUpstreamMap     []handlerWrapper
 		initLocalZones      []nbdns.CustomZone
 		initSerial          uint64
 		inputSerial         uint64
 		inputUpdate         nbdns.Config
 		shouldFail          bool
-		expectedUpstreamMap registeredHandlerMap
+		expectedUpstreamMap []handlerWrapper
 		expectedLocalQs     []dns.Question
 	}{
 		{
 			name:            "Initial Config Should Succeed",
-			initUpstreamMap: make(registeredHandlerMap),
+			initUpstreamMap: nil,
 			initSerial:      0,
 			inputSerial:     1,
 			inputUpdate: nbdns.Config{
@@ -169,20 +154,17 @@ func TestUpdateDNSServer(t *testing.T) {
 					},
 				},
 			},
-			expectedUpstreamMap: registeredHandlerMap{
-				generateDummyHandler("netbird.io", nameServers).ID(): handlerWrapper{
+			expectedUpstreamMap: []handlerWrapper{
+				{
 					domain:   "netbird.io",
-					handler:  dummyHandler,
 					priority: PriorityUpstream,
 				},
-				dummyHandler.ID(): handlerWrapper{
+				{
 					domain:   "netbird.cloud",
-					handler:  dummyHandler,
 					priority: PriorityLocal,
 				},
-				generateDummyHandler(".", nameServers).ID(): handlerWrapper{
+				{
 					domain:   nbdns.RootZone,
-					handler:  dummyHandler,
 					priority: PriorityDefault,
 				},
 			},
@@ -191,10 +173,10 @@ func TestUpdateDNSServer(t *testing.T) {
 		{
 			name:           "New Config Should Succeed",
 			initLocalZones: []nbdns.CustomZone{{Domain: "netbird.cloud", Records: []nbdns.SimpleRecord{{Name: "netbird.cloud", Type: 1, Class: nbdns.DefaultClass, TTL: 300, RData: "10.0.0.1"}}}},
-			initUpstreamMap: registeredHandlerMap{
-				generateDummyHandler(zoneRecords[0].Name, nameServers).ID(): handlerWrapper{
+			initUpstreamMap: []handlerWrapper{
+				{
 					domain:   "netbird.cloud",
-					handler:  dummyHandler,
+					handler:  &mockHandler{},
 					priority: PriorityUpstream,
 				},
 			},
@@ -215,15 +197,13 @@ func TestUpdateDNSServer(t *testing.T) {
 					},
 				},
 			},
-			expectedUpstreamMap: registeredHandlerMap{
-				generateDummyHandler("netbird.io", nameServers).ID(): handlerWrapper{
+			expectedUpstreamMap: []handlerWrapper{
+				{
 					domain:   "netbird.io",
-					handler:  dummyHandler,
 					priority: PriorityUpstream,
 				},
-				"local-resolver": handlerWrapper{
+				{
 					domain:   "netbird.cloud",
-					handler:  dummyHandler,
 					priority: PriorityLocal,
 				},
 			},
@@ -232,7 +212,7 @@ func TestUpdateDNSServer(t *testing.T) {
 		{
 			name:            "Smaller Config Serial Should Be Skipped",
 			initLocalZones:  []nbdns.CustomZone{},
-			initUpstreamMap: make(registeredHandlerMap),
+			initUpstreamMap: nil,
 			initSerial:      2,
 			inputSerial:     1,
 			shouldFail:      true,
@@ -240,7 +220,7 @@ func TestUpdateDNSServer(t *testing.T) {
 		{
 			name:            "Empty NS Group Domain Or Not Primary Element Should Fail",
 			initLocalZones:  []nbdns.CustomZone{},
-			initUpstreamMap: make(registeredHandlerMap),
+			initUpstreamMap: nil,
 			initSerial:      0,
 			inputSerial:     1,
 			inputUpdate: nbdns.Config{
@@ -262,7 +242,7 @@ func TestUpdateDNSServer(t *testing.T) {
 		{
 			name:            "Invalid NS Group Nameservers list Should Fail",
 			initLocalZones:  []nbdns.CustomZone{},
-			initUpstreamMap: make(registeredHandlerMap),
+			initUpstreamMap: nil,
 			initSerial:      0,
 			inputSerial:     1,
 			inputUpdate: nbdns.Config{
@@ -284,7 +264,7 @@ func TestUpdateDNSServer(t *testing.T) {
 		{
 			name:            "Invalid Custom Zone Records list Should Skip",
 			initLocalZones:  []nbdns.CustomZone{},
-			initUpstreamMap: make(registeredHandlerMap),
+			initUpstreamMap: nil,
 			initSerial:      0,
 			inputSerial:     1,
 			inputUpdate: nbdns.Config{
@@ -301,42 +281,41 @@ func TestUpdateDNSServer(t *testing.T) {
 					},
 				},
 			},
-			expectedUpstreamMap: registeredHandlerMap{generateDummyHandler(".", nameServers).ID(): handlerWrapper{
+			expectedUpstreamMap: []handlerWrapper{{
 				domain:   ".",
-				handler:  dummyHandler,
 				priority: PriorityDefault,
 			}},
 		},
 		{
 			name:           "Empty Config Should Succeed and Clean Maps",
 			initLocalZones: []nbdns.CustomZone{{Domain: "netbird.cloud", Records: []nbdns.SimpleRecord{{Name: "netbird.cloud", Type: int(dns.TypeA), Class: nbdns.DefaultClass, TTL: 300, RData: "10.0.0.1"}}}},
-			initUpstreamMap: registeredHandlerMap{
-				generateDummyHandler(zoneRecords[0].Name, nameServers).ID(): handlerWrapper{
+			initUpstreamMap: []handlerWrapper{
+				{
 					domain:   zoneRecords[0].Name,
-					handler:  dummyHandler,
+					handler:  &mockHandler{},
 					priority: PriorityUpstream,
 				},
 			},
 			initSerial:          0,
 			inputSerial:         1,
 			inputUpdate:         nbdns.Config{ServiceEnable: true},
-			expectedUpstreamMap: make(registeredHandlerMap),
+			expectedUpstreamMap: nil,
 			expectedLocalQs:     []dns.Question{},
 		},
 		{
 			name:           "Disabled Service Should clean map",
 			initLocalZones: []nbdns.CustomZone{{Domain: "netbird.cloud", Records: []nbdns.SimpleRecord{{Name: "netbird.cloud", Type: int(dns.TypeA), Class: nbdns.DefaultClass, TTL: 300, RData: "10.0.0.1"}}}},
-			initUpstreamMap: registeredHandlerMap{
-				generateDummyHandler(zoneRecords[0].Name, nameServers).ID(): handlerWrapper{
+			initUpstreamMap: []handlerWrapper{
+				{
 					domain:   zoneRecords[0].Name,
-					handler:  dummyHandler,
+					handler:  &mockHandler{},
 					priority: PriorityUpstream,
 				},
 			},
 			initSerial:          0,
 			inputSerial:         1,
 			inputUpdate:         nbdns.Config{ServiceEnable: false},
-			expectedUpstreamMap: make(registeredHandlerMap),
+			expectedUpstreamMap: nil,
 			expectedLocalQs:     []dns.Question{},
 		},
 	}
@@ -393,7 +372,7 @@ func TestUpdateDNSServer(t *testing.T) {
 				}
 			}()
 
-			dnsServer.dnsMuxMap = testCase.initUpstreamMap
+			dnsServer.dnsMuxHandlers = testCase.initUpstreamMap
 			dnsServer.localResolver.Update(testCase.initLocalZones)
 			dnsServer.updateSerial = testCase.initSerial
 
@@ -405,14 +384,20 @@ func TestUpdateDNSServer(t *testing.T) {
 				t.Fatalf("update dns server should not fail, got error: %v", err)
 			}
 
-			if len(dnsServer.dnsMuxMap) != len(testCase.expectedUpstreamMap) {
-				t.Fatalf("update upstream failed, map size is different than expected, want %d, got %d", len(testCase.expectedUpstreamMap), len(dnsServer.dnsMuxMap))
+			if len(dnsServer.dnsMuxHandlers) != len(testCase.expectedUpstreamMap) {
+				t.Fatalf("update upstream failed, map size is different than expected, want %d, got %d", len(testCase.expectedUpstreamMap), len(dnsServer.dnsMuxHandlers))
 			}
 
-			for key := range testCase.expectedUpstreamMap {
-				_, found := dnsServer.dnsMuxMap[key]
+			for _, expected := range testCase.expectedUpstreamMap {
+				found := false
+				for _, got := range dnsServer.dnsMuxHandlers {
+					if got.domain == expected.domain && got.priority == expected.priority {
+						found = true
+						break
+					}
+				}
 				if !found {
-					t.Fatalf("update upstream failed, key %s was not found in the dnsMuxMap: %#v", key, dnsServer.dnsMuxMap)
+					t.Fatalf("update upstream failed, handler for domain=%s priority=%d not found in dnsMuxHandlers: %#v", expected.domain, expected.priority, dnsServer.dnsMuxHandlers)
 				}
 			}
 
@@ -512,8 +497,8 @@ func TestDNSFakeResolverHandleUpdates(t *testing.T) {
 		}
 	}()
 
-	dnsServer.dnsMuxMap = registeredHandlerMap{
-		"id1": handlerWrapper{
+	dnsServer.dnsMuxHandlers = []handlerWrapper{
+		{
 			domain:   zoneRecords[0].Name,
 			handler:  &local.Resolver{},
 			priority: PriorityUpstream,
@@ -1029,15 +1014,15 @@ func (m *mockService) RegisterMux(string, dns.Handler) {}
 func (m *mockService) DeregisterMux(string)            {}
 
 func TestDefaultServer_UpdateMux(t *testing.T) {
-	baseMatchHandlers := registeredHandlerMap{
-		"upstream-group1": {
+	baseMatchHandlers := []handlerWrapper{
+		{
 			domain: "example.com",
 			handler: &mockHandler{
 				Id: "upstream-group1",
 			},
 			priority: PriorityUpstream,
 		},
-		"upstream-group2": {
+		{
 			domain: "example.com",
 			handler: &mockHandler{
 				Id: "upstream-group2",
@@ -1046,15 +1031,15 @@ func TestDefaultServer_UpdateMux(t *testing.T) {
 		},
 	}
 
-	baseRootHandlers := registeredHandlerMap{
-		"upstream-root1": {
+	baseRootHandlers := []handlerWrapper{
+		{
 			domain: ".",
 			handler: &mockHandler{
 				Id: "upstream-root1",
 			},
 			priority: PriorityDefault,
 		},
-		"upstream-root2": {
+		{
 			domain: ".",
 			handler: &mockHandler{
 				Id: "upstream-root2",
@@ -1063,22 +1048,22 @@ func TestDefaultServer_UpdateMux(t *testing.T) {
 		},
 	}
 
-	baseMixedHandlers := registeredHandlerMap{
-		"upstream-group1": {
+	baseMixedHandlers := []handlerWrapper{
+		{
 			domain: "example.com",
 			handler: &mockHandler{
 				Id: "upstream-group1",
 			},
 			priority: PriorityUpstream,
 		},
-		"upstream-group2": {
+		{
 			domain: "example.com",
 			handler: &mockHandler{
 				Id: "upstream-group2",
 			},
 			priority: PriorityUpstream - 1,
 		},
-		"upstream-other": {
+		{
 			domain: "other.com",
 			handler: &mockHandler{
 				Id: "upstream-other",
@@ -1089,7 +1074,7 @@ func TestDefaultServer_UpdateMux(t *testing.T) {
 
 	tests := []struct {
 		name             string
-		initialHandlers  registeredHandlerMap
+		initialHandlers  []handlerWrapper
 		updates          []handlerWrapper
 		expectedHandlers map[string]string // map[HandlerID]domain
 		description      string
@@ -1373,32 +1358,38 @@ func TestDefaultServer_UpdateMux(t *testing.T) {
 	for _, tt := range tests {
 		t.Run(tt.name, func(t *testing.T) {
 			server := &DefaultServer{
-				dnsMuxMap:    tt.initialHandlers,
-				handlerChain: NewHandlerChain(),
-				service:      &mockService{},
+				dnsMuxHandlers: tt.initialHandlers,
+				handlerChain:   NewHandlerChain(),
+				service:        &mockService{},
 			}
 
 			// Perform the update
 			server.updateMux(tt.updates)
 
 			// Verify the results
-			assert.Equal(t, len(tt.expectedHandlers), len(server.dnsMuxMap),
+			assert.Equal(t, len(tt.expectedHandlers), len(server.dnsMuxHandlers),
 				"Number of handlers after update doesn't match expected")
 
 			// Check each expected handler
 			for id, expectedDomain := range tt.expectedHandlers {
-				handler, exists := server.dnsMuxMap[types.HandlerID(id)]
-				assert.True(t, exists, "Expected handler %s not found", id)
-				if exists {
-					assert.Equal(t, expectedDomain, handler.domain,
+				var found *handlerWrapper
+				for i := range server.dnsMuxHandlers {
+					if server.dnsMuxHandlers[i].handler.ID() == types.HandlerID(id) {
+						found = &server.dnsMuxHandlers[i]
+						break
+					}
+				}
+				assert.NotNil(t, found, "Expected handler %s not found", id)
+				if found != nil {
+					assert.Equal(t, expectedDomain, found.domain,
 						"Domain mismatch for handler %s", id)
 				}
 			}
 
 			// Verify no unexpected handlers exist
-			for HandlerID := range server.dnsMuxMap {
-				_, expected := tt.expectedHandlers[string(HandlerID)]
-				assert.True(t, expected, "Unexpected handler found: %s", HandlerID)
+			for _, entry := range server.dnsMuxHandlers {
+				_, expected := tt.expectedHandlers[string(entry.handler.ID())]
+				assert.True(t, expected, "Unexpected handler found: %s", entry.handler.ID())
 			}
 
 			// Verify the handlerChain state and order
@@ -1413,7 +1404,7 @@ func TestDefaultServer_UpdateMux(t *testing.T) {
 
 				// Verify handler exists in mux
 				foundInMux := false
-				for _, muxEntry := range server.dnsMuxMap {
+				for _, muxEntry := range server.dnsMuxHandlers {
 					if chainEntry.Handler == muxEntry.handler &&
 						chainEntry.Priority == muxEntry.priority &&
 						chainEntry.Pattern == dns.Fqdn(muxEntry.domain) {
@@ -1422,12 +1413,108 @@ func TestDefaultServer_UpdateMux(t *testing.T) {
 					}
 				}
 				assert.True(t, foundInMux,
-					"Handler in chain not found in dnsMuxMap")
+					"Handler in chain not found in dnsMuxHandlers")
 			}
 		})
 	}
 }
 
+// chainHasPattern reports whether the handler chain holds an entry registered
+// for the given fqdn pattern at the given priority.
+func chainHasPattern(s *DefaultServer, pattern string, priority int) bool {
+	for _, h := range s.handlerChain.handlers {
+		if h.OrigPattern == pattern && h.Priority == priority {
+			return true
+		}
+	}
+	return false
+}
+
+// TestDefaultServer_UpdateMux_SharedHandlerZoneRemoval verifies that updateMux
+// tracks each (handler, domain) registration independently when one handler
+// serves multiple zones. Every custom zone is served by the same handler
+// instance (the local resolver, whose ID is the constant "local-resolver"), so
+// removing one zone must deregister exactly that zone's chain entry and leave
+// the others in place. Tracking registrations by handler ID alone collapses all
+// zones onto one entry, leaving removed zones in the chain to answer
+// authoritatively with no records.
+func TestDefaultServer_UpdateMux_SharedHandlerZoneRemoval(t *testing.T) {
+	// One handler serves every custom zone, mirroring s.localResolver.
+	shared := &mockHandler{Id: "local-resolver"}
+
+	server := &DefaultServer{
+		handlerChain: NewHandlerChain(),
+		service:      &mockService{},
+	}
+
+	// Two custom zones under the same handler. The surviving zone is registered
+	// last, mirroring the management emission order.
+	server.updateMux([]handlerWrapper{
+		{domain: "userzone.test", handler: shared, priority: PriorityLocal},
+		{domain: "peerzone.test", handler: shared, priority: PriorityLocal},
+	})
+
+	require.True(t, chainHasPattern(server, "userzone.test.", PriorityLocal),
+		"userzone.test should be registered after the first update")
+	require.True(t, chainHasPattern(server, "peerzone.test.", PriorityLocal),
+		"peerzone.test should be registered after the first update")
+
+	// Remove one zone, keep the other.
+	server.updateMux([]handlerWrapper{
+		{domain: "peerzone.test", handler: shared, priority: PriorityLocal},
+	})
+
+	assert.True(t, chainHasPattern(server, "peerzone.test.", PriorityLocal),
+		"peerzone.test should remain after removing userzone.test")
+	assert.False(t, chainHasPattern(server, "userzone.test.", PriorityLocal),
+		"userzone.test handler must be deregistered, not leaked in the chain")
+}
+
+// TestDefaultServer_UpdateMux_PreservesLocalResolver verifies that updateMux
+// does not tear down the shared local resolver during reconfiguration. The
+// resolver is a process-lifetime singleton reused across config updates;
+// Stop() cancels its lookup context (breaking external CNAME-target
+// resolution) and clears its records. updateMux must deregister its chain
+// entries without stopping it. Records surviving a teardown update is the
+// observable proxy: Stop() would have cleared them.
+func TestDefaultServer_UpdateMux_PreservesLocalResolver(t *testing.T) {
+	resolver := local.NewResolver()
+	require.NoError(t, resolver.RegisterRecord(nbdns.SimpleRecord{
+		Name:  "peer.netbird.cloud.",
+		Type:  int(dns.TypeA),
+		Class: nbdns.DefaultClass,
+		TTL:   300,
+		RData: "10.0.0.1",
+	}))
+
+	server := &DefaultServer{
+		handlerChain:  NewHandlerChain(),
+		service:       &mockService{},
+		localResolver: resolver,
+	}
+
+	server.updateMux([]handlerWrapper{
+		{domain: "netbird.cloud", handler: resolver, priority: PriorityLocal},
+	})
+
+	// Remove the zone. The resolver must survive so its records and lookup
+	// context stay intact for the next registration.
+	server.updateMux(nil)
+
+	var response *dns.Msg
+	resolver.ServeDNS(&test.MockResponseWriter{
+		WriteMsgFunc: func(m *dns.Msg) error {
+			response = m
+			return nil
+		},
+	}, &dns.Msg{Question: []dns.Question{{Name: "peer.netbird.cloud.", Qtype: dns.TypeA, Qclass: dns.ClassINET}}})
+
+	require.NotNil(t, response, "local resolver should answer after teardown")
+	assert.Equal(t, dns.RcodeSuccess, response.Rcode,
+		"local resolver records must survive teardown; updateMux must not Stop() the shared resolver")
+	assert.NotEmpty(t, response.Answer, "answer should contain the surviving record")
+}
+
 func TestExtraDomains(t *testing.T) {
 	tests := []struct {
 		name                string
@@ -2049,7 +2136,6 @@ func TestBuildUpstreamHandler_MergesGroupsPerDomain(t *testing.T) {
 		localResolver: local.NewResolver(),
 		handlerChain:  NewHandlerChain(),
 		hostManager:   &noopHostConfigurator{},
-		dnsMuxMap:     make(registeredHandlerMap),
 	}
 
 	groups := []*nbdns.NameServerGroup{
@@ -2207,7 +2293,7 @@ func TestEvaluateNSGroupHealth(t *testing.T) {
 	}
 }
 
-// healthStubHandler is a minimal dnsMuxMap entry that exposes a fixed
+// healthStubHandler is a minimal dnsMuxHandlers entry that exposes a fixed
 // UpstreamHealth snapshot, letting tests drive recomputeNSGroupStates
 // without spinning up real handlers.
 type healthStubHandler struct {
@@ -2283,12 +2369,11 @@ func newProjTestFixture(t *testing.T) *projTestFixture {
 		ctx:              context.Background(),
 		wgInterface:      &mocWGIface{},
 		statusRecorder:   recorder,
-		dnsMuxMap:        make(registeredHandlerMap),
 		selectedRoutes:   func() route.HAMap { return fx.selected },
 		activeRoutes:     func() route.HAMap { return fx.active },
 		warningDelayBase: defaultWarningDelayBase,
 	}
-	fx.server.dnsMuxMap["example.com"] = handlerWrapper{domain: "example.com", handler: fx.stub, priority: PriorityUpstream}
+	fx.server.dnsMuxHandlers = []handlerWrapper{{domain: "example.com", handler: fx.stub, priority: PriorityUpstream}}
 
 	fx.server.mux.Lock()
 	fx.server.updateNSGroupStates([]*nbdns.NameServerGroup{fx.group})
@@ -2395,7 +2480,6 @@ func TestProjection_OverlayAddrNoRouteDelaysWarning(t *testing.T) {
 		ctx:              context.Background(),
 		wgInterface:      &mocWGIface{},
 		statusRecorder:   recorder,
-		dnsMuxMap:        make(registeredHandlerMap),
 		selectedRoutes:   func() route.HAMap { return nil },
 		activeRoutes:     func() route.HAMap { return nil },
 		warningDelayBase: 50 * time.Millisecond,
@@ -2407,7 +2491,7 @@ func TestProjection_OverlayAddrNoRouteDelaysWarning(t *testing.T) {
 	stub := &healthStubHandler{health: map[netip.AddrPort]UpstreamHealth{
 		overlayPeer: {LastFail: time.Now(), LastErr: "timeout"},
 	}}
-	server.dnsMuxMap["example.com"] = handlerWrapper{domain: "example.com", handler: stub, priority: PriorityUpstream}
+	server.dnsMuxHandlers = []handlerWrapper{{domain: "example.com", handler: stub, priority: PriorityUpstream}}
 
 	server.mux.Lock()
 	server.updateNSGroupStates([]*nbdns.NameServerGroup{group})
@@ -2444,7 +2528,6 @@ func TestProjection_StopClearsHealthState(t *testing.T) {
 		service:           NewServiceViaMemory(wgIface),
 		hostManager:       &noopHostConfigurator{},
 		extraDomains:      map[domain.Domain]int{},
-		dnsMuxMap:         make(registeredHandlerMap),
 		statusRecorder:    peer.NewRecorder("mgm"),
 		selectedRoutes:    func() route.HAMap { return nil },
 		activeRoutes:      func() route.HAMap { return nil },
@@ -2459,7 +2542,7 @@ func TestProjection_StopClearsHealthState(t *testing.T) {
 		NameServers: []nbdns.NameServer{{IP: srv.Addr(), NSType: nbdns.UDPNameServerType, Port: int(srv.Port())}},
 	}
 	stub := &healthStubHandler{health: map[netip.AddrPort]UpstreamHealth{srv: {LastOk: time.Now()}}}
-	server.dnsMuxMap["example.com"] = handlerWrapper{domain: "example.com", handler: stub, priority: PriorityUpstream}
+	server.dnsMuxHandlers = []handlerWrapper{{domain: "example.com", handler: stub, priority: PriorityUpstream}}
 
 	server.mux.Lock()
 	server.updateNSGroupStates([]*nbdns.NameServerGroup{group})
@@ -2484,6 +2567,32 @@ func TestProjection_StopClearsHealthState(t *testing.T) {
 // rule 3: startup failures while the peer is handshaking, then the peer
 // comes up and a query succeeds before the grace window elapses. No
 // warning should ever have fired, and no recovery either.
+func TestWarningDelayBaseFromEnv(t *testing.T) {
+	tests := []struct {
+		name string
+		set  bool
+		val  string
+		want time.Duration
+	}{
+		{name: "unset uses default", set: false, want: defaultWarningDelayBase},
+		{name: "valid override", set: true, val: "90s", want: 90 * time.Second},
+		{name: "valid minutes", set: true, val: "2m", want: 2 * time.Minute},
+		{name: "invalid falls back", set: true, val: "notaduration", want: defaultWarningDelayBase},
+		{name: "zero falls back", set: true, val: "0s", want: defaultWarningDelayBase},
+		{name: "negative falls back", set: true, val: "-30s", want: defaultWarningDelayBase},
+	}
+
+	for _, tc := range tests {
+		t.Run(tc.name, func(t *testing.T) {
+			t.Setenv(envWarningDelay, tc.val)
+			if !tc.set {
+				os.Unsetenv(envWarningDelay)
+			}
+			assert.Equal(t, tc.want, warningDelayBaseFromEnv(), "grace window base")
+		})
+	}
+}
+
 func TestProjection_OverlayRecoversDuringGrace(t *testing.T) {
 	fx := newProjTestFixture(t)
 	fx.server.warningDelayBase = 200 * time.Millisecond
@@ -2595,7 +2704,6 @@ func TestProjection_MixedGroupEmitsImmediately(t *testing.T) {
 	server := &DefaultServer{
 		ctx:              context.Background(),
 		statusRecorder:   recorder,
-		dnsMuxMap:        make(registeredHandlerMap),
 		selectedRoutes:   func() route.HAMap { return overlayMap },
 		activeRoutes:     func() route.HAMap { return nil },
 		warningDelayBase: time.Hour,
@@ -2613,7 +2721,7 @@ func TestProjection_MixedGroupEmitsImmediately(t *testing.T) {
 			overlay: {LastFail: time.Now(), LastErr: "timeout"},
 		},
 	}
-	server.dnsMuxMap["example.com"] = handlerWrapper{domain: "example.com", handler: stub, priority: PriorityUpstream}
+	server.dnsMuxHandlers = []handlerWrapper{{domain: "example.com", handler: stub, priority: PriorityUpstream}}
 
 	server.mux.Lock()
 	server.updateNSGroupStates([]*nbdns.NameServerGroup{group})
@@ -2640,7 +2748,6 @@ func TestDNSLoopPrevention(t *testing.T) {
 		localResolver: local.NewResolver(),
 		handlerChain:  NewHandlerChain(),
 		hostManager:   &noopHostConfigurator{},
-		dnsMuxMap:     make(registeredHandlerMap),
 	}
 
 	tests := []struct {

client/internal/dns/upstream.go

@@ -443,29 +443,32 @@ func (u *upstreamResolverBase) queryUpstream(parentCtx context.Context, r *dns.M
 		return raceResult{}, &upstreamFailure{upstream: upstream, reason: "no response"}
 	}
 
+	// A valid response means the upstream is reachable, whatever the Rcode.
+	u.markUpstreamOk(upstream)
+
 	proto := ""
 	if upstreamProto != nil {
 		proto = upstreamProto.protocol
 	}
 
 	if rm.Rcode == dns.RcodeServerFailure || rm.Rcode == dns.RcodeRefused {
+		// SERVFAIL and REFUSED are per-question outcomes (DNSSEC-bogus names,
+		// refused zones, transient recursion errors), not reachability
+		// problems: fail over for a better answer but keep the upstream healthy.
 		if code, ok := nonRetryableEDE(rm); ok {
 			if !hadEdns {
-				stripOPT(rm)
+				resutil.StripOPT(rm)
 			}
-			u.markUpstreamOk(upstream)
 			return raceResult{msg: rm, upstream: upstream, protocol: proto, ede: edeName(code)}, nil
 		}
 		reason := dns.RcodeToString[rm.Rcode]
-		u.markUpstreamFail(upstream, reason)
 		return raceResult{}, &upstreamFailure{upstream: upstream, reason: reason}
 	}
 
 	if !hadEdns {
-		stripOPT(rm)
+		resutil.StripOPT(rm)
 	}
 
-	u.markUpstreamOk(upstream)
 	return raceResult{msg: rm, upstream: upstream, protocol: proto}, nil
 }
 
@@ -520,22 +523,6 @@ func upstreamUDPSize() uint16 {
 	return dns.MinMsgSize
 }
 
-// stripOPT removes any OPT pseudo-RRs from the response's Extra section so
-// the response complies with RFC 6891 when the client did not advertise EDNS0.
-func stripOPT(rm *dns.Msg) {
-	if len(rm.Extra) == 0 {
-		return
-	}
-	out := rm.Extra[:0]
-	for _, rr := range rm.Extra {
-		if _, ok := rr.(*dns.OPT); ok {
-			continue
-		}
-		out = append(out, rr)
-	}
-	rm.Extra = out
-}
-
 func (u *upstreamResolverBase) handleUpstreamError(err error, upstream netip.AddrPort, startTime time.Time) *upstreamFailure {
 	if !errors.Is(err, context.DeadlineExceeded) && !isTimeout(err) {
 		return &upstreamFailure{upstream: upstream, reason: err.Error()}

client/internal/dns/upstream_test.go

@@ -517,6 +517,78 @@ func TestUpstreamResolver_HealthTracking(t *testing.T) {
 	assert.NotContains(t, health, bad, "sibling upstream should not be queried when primary answers")
 }
 
+// TestUpstreamResolver_HealthTracking_ResponseMeansReachable verifies that an
+// upstream which answers with SERVFAIL or REFUSED is recorded as healthy:
+// those are per-question outcomes from a reachable server and must not mark
+// the upstream unhealthy. Only transport failures (timeouts) do.
+func TestUpstreamResolver_HealthTracking_ResponseMeansReachable(t *testing.T) {
+	a := netip.MustParseAddrPort("192.0.2.10:53")
+	b := netip.MustParseAddrPort("192.0.2.11:53")
+	timeoutErr := &net.OpError{Op: "read", Err: fmt.Errorf("i/o timeout")}
+
+	tests := []struct {
+		name        string
+		respA       mockUpstreamResponse
+		respB       mockUpstreamResponse
+		wantHealthy bool
+	}{
+		{
+			name:        "both SERVFAIL are reachable",
+			respA:       mockUpstreamResponse{msg: buildMockResponse(dns.RcodeServerFailure, "")},
+			respB:       mockUpstreamResponse{msg: buildMockResponse(dns.RcodeServerFailure, "")},
+			wantHealthy: true,
+		},
+		{
+			name:        "both REFUSED are reachable",
+			respA:       mockUpstreamResponse{msg: buildMockResponse(dns.RcodeRefused, "")},
+			respB:       mockUpstreamResponse{msg: buildMockResponse(dns.RcodeRefused, "")},
+			wantHealthy: true,
+		},
+		{
+			name:        "timeout marks unhealthy",
+			respA:       mockUpstreamResponse{err: timeoutErr},
+			respB:       mockUpstreamResponse{err: timeoutErr},
+			wantHealthy: false,
+		},
+	}
+
+	for _, tc := range tests {
+		t.Run(tc.name, func(t *testing.T) {
+			mockClient := &mockUpstreamResolverPerServer{
+				responses: map[string]mockUpstreamResponse{
+					a.String(): tc.respA,
+					b.String(): tc.respB,
+				},
+				rtt: time.Millisecond,
+			}
+
+			ctx, cancel := context.WithCancel(context.Background())
+			defer cancel()
+
+			resolver := &upstreamResolverBase{
+				ctx:             ctx,
+				upstreamClient:  mockClient,
+				upstreamTimeout: UpstreamTimeout,
+			}
+			resolver.addRace([]netip.AddrPort{a, b})
+
+			responseWriter := &test.MockResponseWriter{WriteMsgFunc: func(m *dns.Msg) error { return nil }}
+			resolver.ServeDNS(responseWriter, new(dns.Msg).SetQuestion("example.com.", dns.TypeA))
+
+			health := resolver.UpstreamHealth()
+			require.Contains(t, health, a, "primary upstream should have a health record")
+			if tc.wantHealthy {
+				assert.False(t, health[a].LastOk.IsZero(), "responding upstream should have LastOk set")
+				assert.True(t, health[a].LastFail.IsZero(), "responding upstream should not be marked failed")
+				assert.Empty(t, health[a].LastErr, "responding upstream should have no error")
+			} else {
+				assert.False(t, health[a].LastFail.IsZero(), "timed-out upstream should be marked failed")
+				assert.NotEmpty(t, health[a].LastErr, "timed-out upstream should record an error")
+			}
+		})
+	}
+}
+
 func TestFormatFailures(t *testing.T) {
 	testCases := []struct {
 		name     string
@@ -913,19 +985,6 @@ func TestEDEName(t *testing.T) {
 	assert.Equal(t, "EDE 9999", edeName(9999), "unknown code falls back to numeric")
 }
 
-func TestStripOPT(t *testing.T) {
-	rm := &dns.Msg{
-		Extra: []dns.RR{
-			&dns.OPT{Hdr: dns.RR_Header{Name: ".", Rrtype: dns.TypeOPT}},
-			&dns.A{Hdr: dns.RR_Header{Name: "x.", Rrtype: dns.TypeA}, A: net.IPv4(1, 2, 3, 4)},
-		},
-	}
-	stripOPT(rm)
-	assert.Len(t, rm.Extra, 1, "OPT should be removed, A kept")
-	_, isOPT := rm.Extra[0].(*dns.OPT)
-	assert.False(t, isOPT, "remaining record must not be OPT")
-}
-
 func TestUpstreamResolver_NonRetryableEDEShortCircuits(t *testing.T) {
 	upstream1 := netip.MustParseAddrPort("192.0.2.1:53")
 	upstream2 := netip.MustParseAddrPort("192.0.2.2:53")

client/internal/dnsfwd/forwarder.go

@@ -26,6 +26,15 @@ import (
 const errResolveFailed = "failed to resolve query for domain=%s: %v"
 const upstreamTimeout = 15 * time.Second
 
+// EDE info codes the forwarder emits on upstream failures so the querying
+// client can see the reason without inspecting this peer's logs. They live in
+// the RFC 8914 Private Use range (49152-65535); the Go resolver never exposes a
+// real upstream EDE here, so these cannot collide with a genuine code.
+const (
+	edeNetbirdUpstreamTimeout uint16 = 49152
+	edeNetbirdUpstreamFailure uint16 = 49153
+)
+
 type resolver interface {
 	LookupNetIP(ctx context.Context, network, host string) ([]netip.Addr, error)
 }
@@ -220,7 +229,7 @@ func (f *DNSForwarder) handleDNSQuery(logger *log.Entry, w dns.ResponseWriter, q
 
 	result := resutil.LookupIP(ctx, f.resolver, network, qname, question.Qtype)
 	if result.Err != nil {
-		f.handleDNSError(ctx, logger, w, question, resp, qname, result, startTime)
+		f.handleDNSError(ctx, logger, w, question, resp, qname, result, query.IsEdns0() != nil, startTime)
 		return
 	}
 
@@ -333,6 +342,7 @@ func (f *DNSForwarder) handleDNSError(
 	resp *dns.Msg,
 	domain string,
 	result resutil.LookupResult,
+	reqHasEdns bool,
 	startTime time.Time,
 ) {
 	qType := question.Qtype
@@ -374,6 +384,10 @@ func (f *DNSForwarder) handleDNSError(
 		logger.Warnf(errResolveFailed, domain, result.Err)
 	}
 
+	if reqHasEdns {
+		attachEDE(resp, edeCodeFor(dnsErr), edeText(dnsErr))
+	}
+
 	f.writeResponse(logger, w, resp, domain, startTime)
 }
 
@@ -414,3 +428,33 @@ func (f *DNSForwarder) getMatchingEntries(domain string) (route.ResID, []*Forwar
 
 	return selectedResId, matches
 }
+
+// edeCodeFor maps an upstream lookup error to the NetBird EDE info code.
+func edeCodeFor(dnsErr *net.DNSError) uint16 {
+	if dnsErr != nil && dnsErr.IsTimeout {
+		return edeNetbirdUpstreamTimeout
+	}
+	return edeNetbirdUpstreamFailure
+}
+
+// edeText builds the EDE extra-text describing the class of upstream failure.
+// It deliberately omits the upstream server address, which may be an internal
+// resolver and is exposed to any client permitted to use the route; the full
+// detail stays in the forwarder's local log.
+func edeText(dnsErr *net.DNSError) string {
+	if dnsErr != nil && dnsErr.IsTimeout {
+		return "netbird forwarder: upstream timeout"
+	}
+	return "netbird forwarder: upstream failure"
+}
+
+// attachEDE adds an Extended DNS Error (RFC 8914) option to the response,
+// creating the OPT pseudo-record if the response does not already carry one.
+func attachEDE(resp *dns.Msg, code uint16, text string) {
+	opt := resp.IsEdns0()
+	if opt == nil {
+		resp.SetEdns0(dns.DefaultMsgSize, false)
+		opt = resp.IsEdns0()
+	}
+	opt.Option = append(opt.Option, &dns.EDNS0_EDE{InfoCode: code, ExtraText: text})
+}

client/internal/dnsfwd/forwarder_test.go

@@ -16,6 +16,7 @@ import (
 	"github.com/stretchr/testify/require"
 
 	firewall "github.com/netbirdio/netbird/client/firewall/manager"
+	"github.com/netbirdio/netbird/client/internal/dns/resutil"
 	"github.com/netbirdio/netbird/client/internal/dns/test"
 	"github.com/netbirdio/netbird/client/internal/peer"
 	"github.com/netbirdio/netbird/route"
@@ -617,6 +618,85 @@ func TestDNSForwarder_ResponseCodes(t *testing.T) {
 	}
 }
 
+func TestDNSForwarder_UpstreamFailureEDE(t *testing.T) {
+	tests := []struct {
+		name        string
+		lookupErr   error
+		reqEdns     bool
+		wantEDE     bool
+		wantCode    uint16
+		wantTextHas string
+	}{
+		{
+			name:        "timeout with edns0",
+			lookupErr:   &net.DNSError{Err: "i/o timeout", Server: "10.0.0.53:53", IsTimeout: true},
+			reqEdns:     true,
+			wantEDE:     true,
+			wantCode:    edeNetbirdUpstreamTimeout,
+			wantTextHas: "netbird forwarder: upstream timeout",
+		},
+		{
+			name:        "server failure with edns0",
+			lookupErr:   &net.DNSError{Err: "server misbehaving", Server: "10.0.0.53:53"},
+			reqEdns:     true,
+			wantEDE:     true,
+			wantCode:    edeNetbirdUpstreamFailure,
+			wantTextHas: "netbird forwarder: upstream failure",
+		},
+		{
+			name:      "no edns0 in request omits ede",
+			lookupErr: &net.DNSError{Err: "server misbehaving", Server: "10.0.0.53:53"},
+			reqEdns:   false,
+			wantEDE:   false,
+		},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			mockResolver := &MockResolver{}
+			forwarder := NewDNSForwarder(netip.MustParseAddrPort("127.0.0.1:0"), 300, nil, &peer.Status{}, nil)
+			forwarder.resolver = mockResolver
+
+			d, err := domain.FromString("example.com")
+			require.NoError(t, err)
+			forwarder.UpdateDomains([]*ForwarderEntry{{Domain: d, ResID: "test-res"}})
+
+			mockResolver.On("LookupNetIP", mock.Anything, "ip4", "example.com.").
+				Return([]netip.Addr(nil), tt.lookupErr).Once()
+
+			query := &dns.Msg{}
+			query.SetQuestion("example.com.", dns.TypeA)
+			if tt.reqEdns {
+				query.SetEdns0(dns.DefaultMsgSize, false)
+			}
+
+			var writtenResp *dns.Msg
+			mockWriter := &test.MockResponseWriter{
+				WriteMsgFunc: func(m *dns.Msg) error {
+					writtenResp = m
+					return nil
+				},
+			}
+
+			forwarder.handleDNSQuery(log.NewEntry(log.StandardLogger()), mockWriter, query, time.Now())
+			mockResolver.AssertExpectations(t)
+
+			require.NotNil(t, writtenResp, "expected a response")
+			assert.Equal(t, dns.RcodeServerFailure, writtenResp.Rcode, "upstream failure must be SERVFAIL")
+
+			ede, ok := resutil.ExtractEDE(writtenResp)
+			if !tt.wantEDE {
+				assert.False(t, ok, "response must not carry EDE")
+				return
+			}
+			require.True(t, ok, "response must carry EDE")
+			assert.Equal(t, tt.wantCode, ede.InfoCode, "EDE info code")
+			assert.Contains(t, ede.ExtraText, tt.wantTextHas, "EDE extra-text")
+			assert.NotContains(t, ede.ExtraText, "10.0.0.53", "must not leak upstream server address")
+		})
+	}
+}
+
 func TestDNSForwarder_TCPTruncation(t *testing.T) {
 	// Test that large UDP responses are truncated with TC bit set
 	mockResolver := &MockResolver{}

client/internal/engine.go

@@ -88,6 +88,8 @@ const (
 
 var ErrResetConnection = fmt.Errorf("reset connection")
 
+var ErrEngineAlreadyStarted = errors.New("engine already started")
+
 type EngineConfig struct {
 	WgPort      int
 	WgIfaceName string
@@ -201,6 +203,8 @@ type Engine struct {
 	ctx    context.Context
 	cancel context.CancelFunc
 
+	started bool
+
 	wgInterface WGIface
 
 	udpMux *udpmux.UniversalUDPMuxDefault
@@ -288,9 +292,15 @@ func NewEngine(
 	services EngineServices,
 	mobileDep MobileDependency,
 ) *Engine {
+	// The engine is single-use: a fresh instance is built per connection
+	// cycle (see Client.run), so the run context is created once here rather
+	// than in Start.
+	ctx, cancel := context.WithCancel(clientCtx)
 	engine := &Engine{
 		clientCtx:          clientCtx,
 		clientCancel:       clientCancel,
+		ctx:                ctx,
+		cancel:             cancel,
 		signal:             services.SignalClient,
 		signaler:           peer.NewSignaler(services.SignalClient, config.WgPrivateKey),
 		mgmClient:          services.MgmClient,
@@ -323,8 +333,34 @@ func (e *Engine) Stop() error {
 		log.Debugf("tried stopping engine that is nil")
 		return nil
 	}
+	e.cancel()
 	e.syncMsgMux.Lock()
 
+	e.stopLocked()
+
+	e.syncMsgMux.Unlock()
+
+	timeout := e.calculateShutdownTimeout()
+	log.Debugf("waiting for goroutines to finish with timeout: %v", timeout)
+	shutdownCtx, cancel := context.WithTimeout(context.Background(), timeout)
+	defer cancel()
+
+	if err := waitWithContext(shutdownCtx, &e.shutdownWg); err != nil {
+		log.Warnf("shutdown timeout exceeded after %v, some goroutines may still be running", timeout)
+	}
+
+	log.Infof("stopped Netbird Engine")
+
+	return nil
+}
+
+// stopLocked tears down everything Start may have brought up, in the order
+// teardown requires (DNS before the interface goes down, flow manager after).
+// The caller must hold syncMsgMux. It is shared by Stop and by Start's failure
+// path, so a partially-initialized engine is cleaned up the same way; every
+// step is nil-guarded. It does not wait on shutdownWg — the caller does that
+// after releasing the lock, since the goroutines also take syncMsgMux.
+func (e *Engine) stopLocked() {
 	if e.connMgr != nil {
 		e.connMgr.Close()
 	}
@@ -375,10 +411,6 @@ func (e *Engine) Stop() error {
 	// so dbus and friends don't complain because of a missing interface
 	e.stopDNSServer()
 
-	if e.cancel != nil {
-		e.cancel()
-	}
-
 	e.jobExecutorWG.Wait() // block until job goroutines finish
 
 	e.close()
@@ -397,21 +429,6 @@ func (e *Engine) Stop() error {
 	if err := e.stateManager.PersistState(context.Background()); err != nil {
 		log.Errorf("failed to persist state: %v", err)
 	}
-
-	e.syncMsgMux.Unlock()
-
-	timeout := e.calculateShutdownTimeout()
-	log.Debugf("waiting for goroutines to finish with timeout: %v", timeout)
-	shutdownCtx, cancel := context.WithTimeout(context.Background(), timeout)
-	defer cancel()
-
-	if err := waitWithContext(shutdownCtx, &e.shutdownWg); err != nil {
-		log.Warnf("shutdown timeout exceeded after %v, some goroutines may still be running", timeout)
-	}
-
-	log.Infof("stopped Netbird Engine")
-
-	return nil
 }
 
 // calculateShutdownTimeout returns shutdown timeout: 10s base + 100ms per peer, capped at 30s.
@@ -449,18 +466,38 @@ func waitWithContext(ctx context.Context, wg *sync.WaitGroup) error {
 // Start creates a new WireGuard tunnel interface and listens to events from Signal and Management services
 // Connections to remote peers are not established here.
 // However, they will be established once an event with a list of peers to connect to will be received from Management Service
-func (e *Engine) Start(netbirdConfig *mgmProto.NetbirdConfig, mgmtURL *url.URL) error {
+func (e *Engine) Start(netbirdConfig *mgmProto.NetbirdConfig, mgmtURL *url.URL) (err error) {
 	e.syncMsgMux.Lock()
 	defer e.syncMsgMux.Unlock()
 
-	if err := iface.ValidateMTU(e.config.MTU); err != nil {
+	// The engine is single-use. Reject a duplicate start and a start on an
+	// already-stopped engine (run context cancelled).
+	if e.started {
+		return ErrEngineAlreadyStarted
+	}
+
+	if ctxErr := e.ctx.Err(); ctxErr != nil {
+		return fmt.Errorf("engine already stopped: %w", ctxErr)
+	}
+
+	e.started = true
+
+	// Tear down any partially-initialized state on a failed start. Cancel the
+	// run context first so goroutines started before the failure (connMgr,
+	// srWatcher, monitors) unwind, then stopLocked mirrors Stop's teardown (we
+	// already hold syncMsgMux), cleaning up route/DNS/flow/state managers too,
+	// not just what close() covers.
+	defer func() {
+		if err != nil {
+			e.cancel()
+			e.stopLocked()
+		}
+	}()
+
+	if err = iface.ValidateMTU(e.config.MTU); err != nil {
 		return fmt.Errorf("invalid MTU configuration: %w", err)
 	}
 
-	if e.cancel != nil {
-		e.cancel()
-	}
-	e.ctx, e.cancel = context.WithCancel(e.clientCtx)
 	e.exposeManager = expose.NewManager(e.ctx, e.mgmClient)
 
 	wgIface, err := e.newWgIface()
@@ -494,13 +531,11 @@ func (e *Engine) Start(netbirdConfig *mgmProto.NetbirdConfig, mgmtURL *url.URL)
 
 	initialRoutes, dnsConfig, dnsFeatureFlag, err := e.readInitialSettings()
 	if err != nil {
-		e.close()
 		return fmt.Errorf("read initial settings: %w", err)
 	}
 
 	dnsServer, err := e.newDnsServer(dnsConfig)
 	if err != nil {
-		e.close()
 		return fmt.Errorf("create dns server: %w", err)
 	}
 	e.dnsServer = dnsServer
@@ -535,7 +570,6 @@ func (e *Engine) Start(netbirdConfig *mgmProto.NetbirdConfig, mgmtURL *url.URL)
 
 	if err = e.wgInterfaceCreate(); err != nil {
 		log.Errorf("failed creating tunnel interface %s: [%s]", e.config.WgIfaceName, err.Error())
-		e.close()
 		return fmt.Errorf("create wg interface: %w", err)
 	}
 
@@ -544,7 +578,6 @@ func (e *Engine) Start(netbirdConfig *mgmProto.NetbirdConfig, mgmtURL *url.URL)
 	}
 
 	if err := e.createFirewall(); err != nil {
-		e.close()
 		return err
 	}
 
@@ -556,7 +589,6 @@ func (e *Engine) Start(netbirdConfig *mgmProto.NetbirdConfig, mgmtURL *url.URL)
 	e.udpMux, err = e.wgInterface.Up()
 	if err != nil {
 		log.Errorf("failed to pull up wgInterface [%s]: %s", e.wgInterface.Name(), err.Error())
-		e.close()
 		return fmt.Errorf("up wg interface: %w", err)
 	}
 
@@ -581,9 +613,7 @@ func (e *Engine) Start(netbirdConfig *mgmProto.NetbirdConfig, mgmtURL *url.URL)
 		e.acl = acl.NewDefaultManager(e.firewall)
 	}
 
-	err = e.dnsServer.Initialize()
-	if err != nil {
-		e.close()
+	if err := e.dnsServer.Initialize(); err != nil {
 		return fmt.Errorf("initialize dns server: %w", err)
 	}
 
@@ -595,7 +625,9 @@ func (e *Engine) Start(netbirdConfig *mgmProto.NetbirdConfig, mgmtURL *url.URL)
 	e.srWatcher = guard.NewSRWatcher(e.signal, e.relayManager, e.mobileDep.IFaceDiscover, iceCfg)
 	e.srWatcher.Start(peer.IsForceRelayed())
 
-	e.receiveSignalEvents()
+	if err = e.receiveSignalEvents(); err != nil {
+		return err
+	}
 	e.receiveManagementEvents()
 	e.receiveJobEvents()
 
@@ -647,7 +679,6 @@ func (e *Engine) createFirewall() error {
 
 func (e *Engine) initFirewall() error {
 	if err := e.routeManager.SetFirewall(e.firewall); err != nil {
-		e.close()
 		return fmt.Errorf("set firewall: %w", err)
 	}
 
@@ -1762,7 +1793,7 @@ func (e *Engine) createPeerConn(pubKey string, allowedIPs []netip.Prefix, agentV
 }
 
 // receiveSignalEvents connects to the Signal Service event stream to negotiate connection with remote peers
-func (e *Engine) receiveSignalEvents() {
+func (e *Engine) receiveSignalEvents() error {
 	e.shutdownWg.Add(1)
 	go func() {
 		defer e.shutdownWg.Done()
@@ -1778,6 +1809,13 @@ func (e *Engine) receiveSignalEvents() {
 				return e.ctx.Err()
 			}
 
+			// Self-addressed heartbeat: the signal client's receive watchdog
+			// round-trips this through the server to confirm the receive stream
+			// is delivering. Liveness is already recorded before this handler.
+			if msg.GetBody().GetType() == sProto.Body_HEARTBEAT {
+				return nil
+			}
+
 			conn, ok := e.peerStore.PeerConn(msg.Key)
 			if !ok {
 				return fmt.Errorf("wrongly addressed message %s", msg.Key)
@@ -1826,7 +1864,12 @@ func (e *Engine) receiveSignalEvents() {
 		}
 	}()
 
-	e.signal.WaitStreamConnected()
+	// todo: consider to remove this blocker. I do not see benefit to block the Start operations
+	e.signal.WaitStreamConnected(e.ctx)
+	if err := e.ctx.Err(); err != nil {
+		return fmt.Errorf("wait for signal stream: %w", err)
+	}
+	return nil
 }
 
 func (e *Engine) parseNATExternalIPMappings() []string {

client/internal/engine_test.go

@@ -247,7 +247,7 @@ func TestEngine_SSH(t *testing.T) {
 		return
 	}
 
-	ctx, cancel := context.WithCancel(context.Background())
+	ctx, cancel := context.WithCancel(CtxInitState(context.Background()))
 	defer cancel()
 
 	relayMgr := relayClient.NewManager(ctx, nil, key.PublicKey().String(), iface.DefaultMTU)
@@ -426,7 +426,7 @@ func TestEngine_UpdateNetworkMap(t *testing.T) {
 		return
 	}
 
-	ctx, cancel := context.WithCancel(context.Background())
+	ctx, cancel := context.WithCancel(CtxInitState(context.Background()))
 	defer cancel()
 
 	relayMgr := relayClient.NewManager(ctx, nil, key.PublicKey().String(), iface.DefaultMTU)
@@ -638,7 +638,7 @@ func TestEngine_Sync(t *testing.T) {
 		return
 	}
 
-	ctx, cancel := context.WithCancel(context.Background())
+	ctx, cancel := context.WithCancel(CtxInitState(context.Background()))
 	defer cancel()
 
 	// feed updates to Engine via mocked Management client
@@ -817,7 +817,7 @@ func TestEngine_UpdateNetworkMapWithRoutes(t *testing.T) {
 				return
 			}
 
-			ctx, cancel := context.WithCancel(context.Background())
+			ctx, cancel := context.WithCancel(CtxInitState(context.Background()))
 			defer cancel()
 
 			wgIfaceName := fmt.Sprintf("utun%d", 104+n)
@@ -1024,7 +1024,7 @@ func TestEngine_UpdateNetworkMapWithDNSUpdate(t *testing.T) {
 				return
 			}
 
-			ctx, cancel := context.WithCancel(context.Background())
+			ctx, cancel := context.WithCancel(CtxInitState(context.Background()))
 			defer cancel()
 
 			wgIfaceName := fmt.Sprintf("utun%d", 104+n)

client/internal/profilemanager/config.go

@@ -108,6 +108,10 @@ type ConfigInput struct {
 
 // Config Configuration type
 type Config struct {
+	// Name is the human-readable profile name shown in CLI/UI listings.
+	// It is independent of the profile's on-disk filename (which is the ID).
+	Name string
+
 	// Wireguard private key of local peer
 	PrivateKey                    string
 	PreSharedKey                  string
@@ -270,6 +274,16 @@ func createNewConfig(input ConfigInput) (*Config, error) {
 }
 
 func (config *Config) apply(input ConfigInput) (updated bool, err error) {
+	if config.Name != "" {
+		sanitized, err := sanitizeDisplayName(config.Name)
+		if err != nil {
+			return false, fmt.Errorf("invalid profile name: %w", err)
+		}
+		if sanitized != config.Name {
+			config.Name = sanitized
+			updated = true
+		}
+	}
 	if config.ManagementURL == nil {
 		log.Infof("using default Management URL %s", DefaultManagementURL)
 		config.ManagementURL, err = parseURL("Management URL", DefaultManagementURL)

client/internal/profilemanager/id.go

@@ -0,0 +1,118 @@
+package profilemanager
+
+import (
+	"crypto/rand"
+	"encoding/hex"
+	"fmt"
+	"path/filepath"
+	"strings"
+	"unicode"
+	"unicode/utf8"
+)
+
+const (
+	// profileIDByteLen is the number of random bytes generated for a new
+	// profile ID. The resulting hex string is twice this length.
+	profileIDByteLen = 16
+
+	// shortIDLen is the number of leading characters of an ID we render in
+	// list output. Profiles per device are few, so 8 chars is collision-safe
+	// in practice and easy to type as a prefix.
+	shortIDLen = 8
+
+	// maxProfileNameLen caps the human-readable profile name to keep table
+	// output legible and prevent denial-of-service via huge JSON fields.
+	maxProfileNameLen = 128
+
+	// maxProfileIDLen bounds the on-disk filename we'll accept. New
+	// IDs are 32 hex chars, legacy stems are sanitized profile names. The
+	// cap is generous enough to cover both without permitting absurdly
+	// long filenames.
+	maxProfileIDLen = 64
+)
+
+type ID string
+
+// generateProfileID returns a new random hex ID for a profile file.
+func generateProfileID() (ID, error) {
+	buf := make([]byte, profileIDByteLen)
+	if _, err := rand.Read(buf); err != nil {
+		return "", fmt.Errorf("read random bytes: %w", err)
+	}
+	return ID(hex.EncodeToString(buf)), nil
+}
+
+// IsValidProfileFilenameStem reports whether id is safe to use as the stem
+// of a profile JSON filename.
+func IsValidProfileFilenameStem(id ID) bool {
+	s := id.String()
+	if s == "" || len(s) > maxProfileIDLen {
+		return false
+	}
+	if s == defaultProfileName {
+		return true
+	}
+	if strings.ContainsAny(s, `/\`) || strings.Contains(s, "..") {
+		return false
+	}
+	// filepath.Base catches any leftover separators on platforms with
+	// exotic path conventions.
+	if filepath.Base(s) != s {
+		return false
+	}
+	for _, r := range s {
+		if !(unicode.IsLetter(r) || unicode.IsDigit(r) || r == '_' || r == '-') {
+			return false
+		}
+	}
+	return true
+}
+
+// sanitizeDisplayName normalizes a user-supplied profile display name for
+// storage. It strips ASCII control characters, rejects invalid UTF-8, and
+// caps the length. Emojis, spaces, punctuation, and non-ASCII letters are
+// preserved. Returns an error if nothing usable remains.
+func sanitizeDisplayName(name string) (string, error) {
+	if !utf8.ValidString(name) {
+		return "", fmt.Errorf("name is not valid UTF-8")
+	}
+	name = StripCtrlChars(name)
+	name = strings.TrimSpace(name)
+	if name == "" {
+		return "", fmt.Errorf("name is empty after sanitization")
+	}
+	if utf8.RuneCountInString(name) > maxProfileNameLen {
+		return "", fmt.Errorf("name exceeds %d characters", maxProfileNameLen)
+	}
+	return name, nil
+}
+
+// StripCtrlChars control characters from a name before printing it.
+func StripCtrlChars(name string) string {
+	var b strings.Builder
+	b.Grow(len(name))
+	for _, r := range name {
+		// Skip C0 controls and DEL, plus C1 controls (0x80–0x9F).
+		if r < 0x20 || r == 0x7F || (r >= 0x80 && r <= 0x9F) {
+			continue
+		}
+		b.WriteRune(r)
+	}
+	return b.String()
+}
+
+// ShortID truncates an ID for display.
+func (id ID) ShortID() string {
+	if id == DefaultProfileName {
+		return DefaultProfileName
+	}
+	runes := []rune(id)
+	if len(runes) <= shortIDLen {
+		return id.String()
+	}
+	return string(runes[:shortIDLen])
+}
+
+func (id ID) String() string {
+	return string(id)
+}

client/internal/profilemanager/profilemanager.go

@@ -19,19 +19,41 @@ const (
 )
 
 type Profile struct {
-	Name     string
+	// ID is the on-disk filename stem (without .json). For new profiles
+	// it is a 32-char hex string; legacy profiles created before the
+	// ID-keyed layout keep their original name as their ID. The reserved
+	// value "default" identifies the special default profile.
+	ID ID
+	// Name is the human-readable display name. Falls back to ID when the
+	// underlying JSON has no "name" field set.
+	Name string
+	// Path is the absolute path to the profile JSON. Populated by the
+	// loader so callers do not have to reconstruct it from ID + dir.
+	Path     string
 	IsActive bool
 }
 
 func (p *Profile) FilePath() (string, error) {
-	if p.Name == "" {
-		return "", fmt.Errorf("active profile name is empty")
+	if p.Path != "" {
+		return p.Path, nil
 	}
 
-	if p.Name == defaultProfileName {
+	id := p.ID
+	if id == "" {
+		id = ID(p.Name)
+	}
+	if id == "" {
+		return "", fmt.Errorf("profile ID is empty")
+	}
+
+	if id == defaultProfileName {
 		return DefaultConfigPath, nil
 	}
 
+	if !IsValidProfileFilenameStem(id) {
+		return "", fmt.Errorf("invalid profile ID: %q", id)
+	}
+
 	username, err := user.Current()
 	if err != nil {
 		return "", fmt.Errorf("failed to get current user: %w", err)
@@ -42,10 +64,13 @@ func (p *Profile) FilePath() (string, error) {
 		return "", fmt.Errorf("failed to get config directory for user %s: %w", username.Username, err)
 	}
 
-	return filepath.Join(configDir, p.Name+".json"), nil
+	return filepath.Join(configDir, id.String()+".json"), nil
 }
 
 func (p *Profile) IsDefault() bool {
+	if p.ID != "" {
+		return p.ID == defaultProfileName
+	}
 	return p.Name == defaultProfileName
 }
 
@@ -57,18 +82,24 @@ func NewProfileManager() *ProfileManager {
 	return &ProfileManager{}
 }
 
+// GetActiveProfile returns the active profile as recorded in the local
+// user state file. Only ID is populated.
 func (pm *ProfileManager) GetActiveProfile() (*Profile, error) {
 	pm.mu.Lock()
 	defer pm.mu.Unlock()
 
-	prof := pm.getActiveProfileState()
-	return &Profile{Name: prof}, nil
+	id := pm.getActiveProfileState()
+	return &Profile{ID: id}, nil
 }
 
-func (pm *ProfileManager) SwitchProfile(profileName string) error {
-	profileName = sanitizeProfileName(profileName)
+// SwitchProfile records the given profile ID as active in the local user
+// state file.
+func (pm *ProfileManager) SwitchProfile(id ID) error {
+	if id != defaultProfileName && !IsValidProfileFilenameStem(id) {
+		return fmt.Errorf("invalid profile ID: %q", id)
+	}
 
-	if err := pm.setActiveProfileState(profileName); err != nil {
+	if err := pm.setActiveProfileState(id); err != nil {
 		return fmt.Errorf("failed to switch profile: %w", err)
 	}
 	return nil
@@ -85,7 +116,7 @@ func sanitizeProfileName(name string) string {
 	}, name)
 }
 
-func (pm *ProfileManager) getActiveProfileState() string {
+func (pm *ProfileManager) getActiveProfileState() ID {
 
 	configDir, err := getConfigDir()
 	if err != nil {
@@ -113,10 +144,10 @@ func (pm *ProfileManager) getActiveProfileState() string {
 		return defaultProfileName
 	}
 
-	return profileName
+	return ID(profileName)
 }
 
-func (pm *ProfileManager) setActiveProfileState(profileName string) error {
+func (pm *ProfileManager) setActiveProfileState(id ID) error {
 
 	configDir, err := getConfigDir()
 	if err != nil {
@@ -125,7 +156,7 @@ func (pm *ProfileManager) setActiveProfileState(profileName string) error {
 
 	statePath := filepath.Join(configDir, activeProfileStateFilename)
 
-	err = os.WriteFile(statePath, []byte(profileName), 0600)
+	err = os.WriteFile(statePath, []byte(id), 0600)
 	if err != nil {
 		return fmt.Errorf("failed to write active profile state: %w", err)
 	}
@@ -142,7 +173,7 @@ func GetLoginHint() string {
 		return ""
 	}
 
-	profileState, err := pm.GetProfileState(activeProf.Name)
+	profileState, err := pm.GetProfileState(activeProf.ID)
 	if err != nil {
 		log.Debugf("failed to get profile state for login hint: %v", err)
 		return ""

client/internal/profilemanager/profilemanager_test.go

@@ -50,14 +50,14 @@ func TestServiceManager_CreateAndGetDefaultProfile(t *testing.T) {
 
 			state, err := sm.GetActiveProfileState()
 			assert.NoError(t, err)
-			assert.Equal(t, state.Name, defaultProfileName) // No active profile state yet
+			assert.Equal(t, defaultProfileName, state.ID.String()) // No active profile state yet
 
 			err = sm.SetActiveProfileStateToDefault()
 			assert.NoError(t, err)
 
 			active, err := sm.GetActiveProfileState()
 			assert.NoError(t, err)
-			assert.Equal(t, "default", active.Name)
+			assert.Equal(t, "default", active.ID.String())
 		})
 	})
 }
@@ -92,14 +92,14 @@ func TestServiceManager_SetActiveProfileState(t *testing.T) {
 			currUser, err := user.Current()
 			assert.NoError(t, err)
 			sm := &ServiceManager{}
-			state := &ActiveProfileState{Name: "foo", Username: currUser.Username}
+			state := &ActiveProfileState{ID: "foo", Username: currUser.Username}
 			err = sm.SetActiveProfileState(state)
 			assert.NoError(t, err)
 
 			// Should error on nil or incomplete state
 			err = sm.SetActiveProfileState(nil)
 			assert.Error(t, err)
-			err = sm.SetActiveProfileState(&ActiveProfileState{Name: "", Username: ""})
+			err = sm.SetActiveProfileState(&ActiveProfileState{ID: "", Username: ""})
 			assert.Error(t, err)
 		})
 	})

client/internal/profilemanager/service.go

@@ -2,6 +2,7 @@ package profilemanager
 
 import (
 	"context"
+	"encoding/json"
 	"errors"
 	"fmt"
 	"io"
@@ -23,12 +24,43 @@ var (
 	DefaultConfigPathDir   = ""
 	DefaultConfigPath      = ""
 	ActiveProfileStatePath = ""
-)
 
-var (
 	ErrorOldDefaultConfigNotFound = errors.New("old default config not found")
 )
 
+// ErrAmbiguousHandle is returned when a profile handle (ID prefix or name)
+// matches more than one profile. Callers can render Candidates to help the
+// user disambiguate.
+type ErrAmbiguousHandle struct {
+	Handle     string
+	Candidates []Profile
+	Kind       AmbiguityKind
+}
+
+// AmbiguityKind describes which matcher produced the ambiguity, so callers
+// can tailor the error message.
+type AmbiguityKind int
+
+const (
+	AmbiguityKindIDPrefix AmbiguityKind = iota
+	AmbiguityKindName
+)
+
+// profileMeta is the minimal slice of a profile JSON we need, so we avoid
+// reading all fields
+type profileMeta struct {
+	Name string
+}
+
+func (e *ErrAmbiguousHandle) Error() string {
+	switch e.Kind {
+	case AmbiguityKindIDPrefix:
+		return fmt.Sprintf("ID prefix %q is ambiguous (matches %d profiles)", e.Handle, len(e.Candidates))
+	default:
+		return fmt.Sprintf("name %q is ambiguous (%d profiles share this name)", e.Handle, len(e.Candidates))
+	}
+}
+
 func init() {
 
 	DefaultConfigPathDir = "/var/lib/netbird/"
@@ -54,25 +86,34 @@ func init() {
 }
 
 type ActiveProfileState struct {
-	Name     string `json:"name"`
+	// ID is the on-disk filename stem of the active profile. The JSON tag stays
+	// as "name" for backwards compatibility with active state files written
+	// before the ID-based config files. Legacy values were profile names, which
+	// were also the legacy filename stems, so they still resolve to the correct
+	// file on disk.
+	ID       ID     `json:"name"`
 	Username string `json:"username"`
 }
 
 func (a *ActiveProfileState) FilePath() (string, error) {
-	if a.Name == "" {
-		return "", fmt.Errorf("active profile name is empty")
+	if a.ID == "" {
+		return "", fmt.Errorf("active profile ID is empty")
 	}
 
-	if a.Name == defaultProfileName {
+	if a.ID == defaultProfileName {
 		return DefaultConfigPath, nil
 	}
 
+	if !IsValidProfileFilenameStem(a.ID) {
+		return "", fmt.Errorf("invalid profile ID: %q", a.ID)
+	}
+
 	configDir, err := getConfigDirForUser(a.Username)
 	if err != nil {
 		return "", fmt.Errorf("failed to get config directory for user %s: %w", a.Username, err)
 	}
 
-	return filepath.Join(configDir, a.Name+".json"), nil
+	return filepath.Join(configDir, a.ID.String()+".json"), nil
 }
 
 type ServiceManager struct {
@@ -178,7 +219,7 @@ func (s *ServiceManager) GetActiveProfileState() (*ActiveProfileState, error) {
 				return nil, fmt.Errorf("failed to set active profile to default: %w", err)
 			}
 			return &ActiveProfileState{
-				Name:     "default",
+				ID:       defaultProfileName,
 				Username: "",
 			}, nil
 		} else {
@@ -186,12 +227,12 @@ func (s *ServiceManager) GetActiveProfileState() (*ActiveProfileState, error) {
 		}
 	}
 
-	if activeProfile.Name == "" {
+	if activeProfile.ID == "" {
 		if err := s.SetActiveProfileStateToDefault(); err != nil {
 			return nil, fmt.Errorf("failed to set active profile to default: %w", err)
 		}
 		return &ActiveProfileState{
-			Name:     "default",
+			ID:       defaultProfileName,
 			Username: "",
 		}, nil
 	}
@@ -216,25 +257,29 @@ func (s *ServiceManager) setDefaultActiveState() error {
 }
 
 func (s *ServiceManager) SetActiveProfileState(a *ActiveProfileState) error {
-	if a == nil || a.Name == "" {
+	if a == nil || a.ID == "" {
 		return errors.New("invalid active profile state")
 	}
 
-	if a.Name != defaultProfileName && a.Username == "" {
-		return fmt.Errorf("username must be set for non-default profiles, got: %s", a.Name)
+	if a.ID != defaultProfileName && a.Username == "" {
+		return fmt.Errorf("username must be set for non-default profiles, got: %s", a.ID)
+	}
+
+	if a.ID != defaultProfileName && !IsValidProfileFilenameStem(a.ID) {
+		return fmt.Errorf("invalid profile ID: %q", a.ID)
 	}
 
 	if err := util.WriteJsonWithRestrictedPermission(context.Background(), ActiveProfileStatePath, a); err != nil {
 		return fmt.Errorf("failed to write active profile state: %w", err)
 	}
 
-	log.Infof("active profile set to %s for %s", a.Name, a.Username)
+	log.Infof("active profile set to %s for %s", a.ID, a.Username)
 	return nil
 }
 
 func (s *ServiceManager) SetActiveProfileStateToDefault() error {
 	return s.SetActiveProfileState(&ActiveProfileState{
-		Name:     "default",
+		ID:       defaultProfileName,
 		Username: "",
 	})
 }
@@ -243,57 +288,117 @@ func (s *ServiceManager) DefaultProfilePath() string {
 	return DefaultConfigPath
 }
 
-func (s *ServiceManager) AddProfile(profileName, username string) error {
+// AddProfile creates a new profile with a generated ID. The user-supplied
+// displayName is stored inside the JSON's name field, the on-disk filename
+// uses the generated ID.
+//
+// The returned Profile carries the freshly-generated ID so callers can
+// show it to the user (and so the gRPC AddProfileResponse can include
+// it).
+func (s *ServiceManager) AddProfile(displayName, username string) (*Profile, error) {
 	configDir, err := s.getConfigDir(username)
 	if err != nil {
-		return fmt.Errorf("failed to get config directory: %w", err)
+		return nil, fmt.Errorf("failed to get config directory: %w", err)
 	}
 
-	profileName = sanitizeProfileName(profileName)
-
-	if profileName == defaultProfileName {
-		return fmt.Errorf("cannot create profile with reserved name: %s", defaultProfileName)
-	}
-
-	profPath := filepath.Join(configDir, profileName+".json")
-	profileExists, err := fileExists(profPath)
+	displayName, err = sanitizeDisplayName(displayName)
 	if err != nil {
-		return fmt.Errorf("failed to check if profile exists: %w", err)
-	}
-	if profileExists {
-		return ErrProfileAlreadyExists
+		return nil, fmt.Errorf("invalid profile name: %w", err)
 	}
 
+	id, err := generateProfileID()
+	if err != nil {
+		return nil, fmt.Errorf("generate profile id: %w", err)
+	}
+
+	profPath := filepath.Join(configDir, id.String()+".json")
 	cfg, err := createNewConfig(ConfigInput{ConfigPath: profPath})
 	if err != nil {
-		return fmt.Errorf("failed to create new config: %w", err)
+		return nil, fmt.Errorf("failed to create new config: %w", err)
+	}
+	cfg.Name = displayName
+
+	if err := util.WriteJson(context.Background(), profPath, cfg); err != nil {
+		return nil, fmt.Errorf("failed to write profile config: %w", err)
 	}
 
-	err = util.WriteJson(context.Background(), profPath, cfg)
+	return &Profile{
+		ID:   id,
+		Name: displayName,
+		Path: profPath,
+	}, nil
+}
+
+func (s *ServiceManager) RenameProfile(id ID, username string, newName string) error {
+	displayName, err := sanitizeDisplayName(newName)
 	if err != nil {
-		return fmt.Errorf("failed to write profile config: %w", err)
+		return fmt.Errorf("invalid profile name: %w", err)
 	}
 
+	if !IsValidProfileFilenameStem(id) {
+		return fmt.Errorf("invalid profile ID: %q", id)
+	}
+
+	profiles, err := s.loadAllProfiles(username)
+	if err != nil {
+		return fmt.Errorf("load profiles: %w", err)
+	}
+
+	var target *Profile
+	for i := range profiles {
+		if profiles[i].ID == id {
+			target = &profiles[i]
+			break
+		}
+	}
+	if target == nil {
+		return ErrProfileNotFound
+	}
+
+	data, err := os.ReadFile(target.Path)
+	if err != nil {
+		return err
+	}
+	var cfg Config
+	if err := json.Unmarshal(data, &cfg); err != nil {
+		return err
+	}
+	cfg.Name = displayName
+
+	if err := util.WriteJson(context.Background(), target.Path, cfg); err != nil {
+		return fmt.Errorf("failed to write profile name: %w", err)
+	}
 	return nil
 }
 
-func (s *ServiceManager) RemoveProfile(profileName, username string) error {
-	configDir, err := s.getConfigDir(username)
-	if err != nil {
-		return fmt.Errorf("failed to get config directory: %w", err)
+// RemoveProfile deletes the profile identified by id. Callers must have
+// already resolved any user-supplied handle to a concrete ID via
+// ResolveProfile.
+func (s *ServiceManager) RemoveProfile(id ID, username string) error {
+	if id == defaultProfileName {
+		defaultName := readProfileName(DefaultConfigPath)
+		if defaultName == "" {
+			defaultName = defaultProfileName
+		}
+		return fmt.Errorf("cannot remove default profile with name: %s", defaultName)
+	}
+	if !IsValidProfileFilenameStem(id) {
+		return fmt.Errorf("invalid profile ID: %q", id)
 	}
 
-	profileName = sanitizeProfileName(profileName)
-
-	if profileName == defaultProfileName {
-		return fmt.Errorf("cannot remove profile with reserved name: %s", defaultProfileName)
-	}
-	profPath := filepath.Join(configDir, profileName+".json")
-	profileExists, err := fileExists(profPath)
+	profiles, err := s.loadAllProfiles(username)
 	if err != nil {
-		return fmt.Errorf("failed to check if profile exists: %w", err)
+		return fmt.Errorf("load profiles: %w", err)
 	}
-	if !profileExists {
+
+	var target *Profile
+	for i := range profiles {
+		if profiles[i].ID == id {
+			target = &profiles[i]
+			break
+		}
+	}
+	if target == nil {
 		return ErrProfileNotFound
 	}
 
@@ -301,57 +406,26 @@ func (s *ServiceManager) RemoveProfile(profileName, username string) error {
 	if err != nil && !errors.Is(err, ErrNoActiveProfile) {
 		return fmt.Errorf("failed to get active profile: %w", err)
 	}
-
-	if activeProf != nil && activeProf.Name == profileName {
-		return fmt.Errorf("cannot remove active profile: %s", profileName)
+	if activeProf != nil && activeProf.ID == id {
+		return fmt.Errorf("cannot remove active profile: %s", id)
 	}
 
-	err = util.RemoveJson(profPath)
-	if err != nil {
+	if err := util.RemoveJson(target.Path); err != nil {
 		return fmt.Errorf("failed to remove profile config: %w", err)
 	}
+
+	stateFile := filepath.Join(filepath.Dir(target.Path), id.String()+".state.json")
+	if err := os.Remove(stateFile); err != nil && !os.IsNotExist(err) {
+		log.Warnf("failed to remove profile state file %s: %v", stateFile, err)
+	}
+
 	return nil
 }
 
+// ListProfiles returns every profile for the given user, including the
+// default profile, with IsActive flags set.
 func (s *ServiceManager) ListProfiles(username string) ([]Profile, error) {
-	configDir, err := s.getConfigDir(username)
-	if err != nil {
-		return nil, fmt.Errorf("failed to get config directory: %w", err)
-	}
-
-	files, err := util.ListFiles(configDir, "*.json")
-	if err != nil {
-		return nil, fmt.Errorf("failed to list profile files: %w", err)
-	}
-
-	var filtered []string
-	for _, file := range files {
-		if strings.HasSuffix(file, "state.json") {
-			continue // skip state files
-		}
-		filtered = append(filtered, file)
-	}
-	sort.Strings(filtered)
-
-	var activeProfName string
-	activeProf, err := s.GetActiveProfileState()
-	if err == nil {
-		activeProfName = activeProf.Name
-	}
-
-	var profiles []Profile
-	// add default profile always
-	profiles = append(profiles, Profile{Name: defaultProfileName, IsActive: activeProfName == "" || activeProfName == defaultProfileName})
-	for _, file := range filtered {
-		profileName := strings.TrimSuffix(filepath.Base(file), ".json")
-		var isActive bool
-		if activeProfName != "" && activeProfName == profileName {
-			isActive = true
-		}
-		profiles = append(profiles, Profile{Name: profileName, IsActive: isActive})
-	}
-
-	return profiles, nil
+	return s.loadAllProfiles(username)
 }
 
 // GetStatePath returns the path to the state file based on the operating system
@@ -369,7 +443,12 @@ func (s *ServiceManager) GetStatePath() string {
 		return defaultStatePath
 	}
 
-	if activeProf.Name == defaultProfileName {
+	if activeProf.ID == defaultProfileName {
+		return defaultStatePath
+	}
+
+	if !IsValidProfileFilenameStem(activeProf.ID) {
+		log.Warnf("invalid active profile ID %q, using default state path", activeProf.ID)
 		return defaultStatePath
 	}
 
@@ -379,7 +458,7 @@ func (s *ServiceManager) GetStatePath() string {
 		return defaultStatePath
 	}
 
-	return filepath.Join(configDir, activeProf.Name+".state.json")
+	return filepath.Join(configDir, activeProf.ID.String()+".state.json")
 }
 
 // getConfigDir returns the profiles directory, using profilesDir if set, otherwise getConfigDirForUser
@@ -390,3 +469,169 @@ func (s *ServiceManager) getConfigDir(username string) (string, error) {
 
 	return getConfigDirForUser(username)
 }
+
+// loadAllProfiles returns every profile visible to the daemon for the
+// given user, including the default profile. The returned slice is sorted
+// by ID for a stable display order.
+//
+// Each Profile is fully populated: ID is the filename stem, Name comes
+// from the JSON's "name" field (falling back to the filename stem when absent)
+// and Path is built from a basename read off disk.
+func (s *ServiceManager) loadAllProfiles(username string) ([]Profile, error) {
+	activeID, activeIsDefault := s.activeProfileID()
+	defaultName := readProfileName(DefaultConfigPath)
+	if defaultName == "" {
+		defaultName = defaultProfileName
+	}
+
+	profiles := []Profile{{
+		ID:       defaultProfileName,
+		Name:     defaultName,
+		Path:     DefaultConfigPath,
+		IsActive: activeIsDefault,
+	}}
+
+	configDir, err := s.getConfigDir(username)
+	if err != nil {
+		return nil, fmt.Errorf("get config directory: %w", err)
+	}
+
+	entries, err := os.ReadDir(configDir)
+	if err != nil {
+		if errors.Is(err, os.ErrNotExist) {
+			return profiles, nil
+		}
+		return nil, fmt.Errorf("read profile directory: %w", err)
+	}
+
+	var fileProfiles []Profile
+	for _, entry := range entries {
+		if entry.IsDir() {
+			continue
+		}
+		base := entry.Name()
+		if !strings.HasSuffix(base, ".json") {
+			continue
+		}
+		if strings.HasSuffix(base, ".state.json") {
+			continue
+		}
+		stem := ID(strings.TrimSuffix(base, ".json"))
+		if stem == defaultProfileName {
+			// default lives at the top-level config dir, not under /<user>
+			continue
+		}
+		if !IsValidProfileFilenameStem(ID(stem)) {
+			continue
+		}
+		path := filepath.Join(configDir, base)
+		name := readProfileName(path)
+		if name == "" {
+			name = stem.String()
+		}
+		fileProfiles = append(fileProfiles, Profile{
+			ID:       stem,
+			Name:     name,
+			Path:     path,
+			IsActive: stem == ID(activeID),
+		})
+	}
+
+	sort.Slice(fileProfiles, func(i, j int) bool {
+		if fileProfiles[i].Name != fileProfiles[j].Name {
+			return fileProfiles[i].Name < fileProfiles[j].Name
+		}
+		// Sort tie-break on ID so duplicate names always render in the same order.
+		return fileProfiles[i].ID < fileProfiles[j].ID
+	})
+	profiles = append(profiles, fileProfiles...)
+	return profiles, nil
+}
+
+// readProfileName parses just the "name" field from the profile Json.
+func readProfileName(path string) string {
+	data, err := os.ReadFile(path)
+	if err != nil {
+		return ""
+	}
+	var meta profileMeta
+	if err := json.Unmarshal(data, &meta); err != nil {
+		return ""
+	}
+	return meta.Name
+}
+
+// activeProfileID returns the currently-active profile's ID. The second
+// return value is true when the active profile is the default one.
+func (s *ServiceManager) activeProfileID() (ID, bool) {
+	state, err := s.GetActiveProfileState()
+	if err != nil || state == nil {
+		return defaultProfileName, true
+	}
+	if state.ID == "" || state.ID == defaultProfileName {
+		return defaultProfileName, true
+	}
+	return state.ID, false
+}
+
+// ResolveProfile turns a user-supplied handle into a Profile. Resolution
+// precedence is: exact ID match, then unique exact name, then unique ID
+// prefix. Ambiguous matches return *ErrAmbiguousHandle so callers can
+// surface the candidates.
+func (s *ServiceManager) ResolveProfile(handle, username string) (*Profile, error) {
+	if handle == "" {
+		return nil, fmt.Errorf("profile handle is empty")
+	}
+
+	profiles, err := s.loadAllProfiles(username)
+	if err != nil {
+		return nil, err
+	}
+
+	for i := range profiles {
+		if profiles[i].ID == ID(handle) {
+			return &profiles[i], nil
+		}
+	}
+
+	var nameMatches []Profile
+	for i := range profiles {
+		if profiles[i].Name == handle {
+			nameMatches = append(nameMatches, profiles[i])
+		}
+	}
+	if len(nameMatches) == 1 {
+		return &nameMatches[0], nil
+	}
+	if len(nameMatches) > 1 {
+		return nil, &ErrAmbiguousHandle{
+			Handle:     handle,
+			Candidates: nameMatches,
+			Kind:       AmbiguityKindName,
+		}
+	}
+
+	// ID prefix match. Skip the default profile so `select d` does not
+	// accidentally pick it via prefix.
+	var prefixMatches []Profile
+	for i := range profiles {
+		if profiles[i].ID == defaultProfileName {
+			continue
+		}
+		if strings.HasPrefix(profiles[i].ID.String(), handle) {
+			prefixMatches = append(prefixMatches, profiles[i])
+		}
+	}
+	if len(prefixMatches) == 1 {
+		return &prefixMatches[0], nil
+	}
+	if len(prefixMatches) > 1 {
+		return nil, &ErrAmbiguousHandle{
+			Handle:     handle,
+			Candidates: prefixMatches,
+			Kind:       AmbiguityKindIDPrefix,
+		}
+	}
+
+	return nil, ErrProfileNotFound
+}

client/internal/profilemanager/service_test.go

@@ -0,0 +1,230 @@
+package profilemanager
+
+import (
+	"context"
+	"errors"
+	"os"
+	"os/user"
+	"path/filepath"
+	"strings"
+	"testing"
+
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+
+	"github.com/netbirdio/netbird/util"
+)
+
+// withTestSM wires up patched globals + a clean config dir and returns a
+// fully initialized ServiceManager plus the username we are scoped to.
+func withTestSM(t *testing.T, fn func(sm *ServiceManager, username string)) {
+	t.Helper()
+	withTempConfigDir(t, func(configDir string) {
+		withPatchedGlobals(t, configDir, func() {
+			u, err := user.Current()
+			require.NoError(t, err)
+			sm := &ServiceManager{}
+			require.NoError(t, sm.CreateDefaultProfile())
+			fn(sm, u.Username)
+		})
+	})
+}
+
+func TestServiceProfile_ExactID(t *testing.T) {
+	withTestSM(t, func(sm *ServiceManager, username string) {
+		created, err := sm.AddProfile("work", username)
+		require.NoError(t, err)
+
+		got, err := sm.ResolveProfile(created.ID.String(), username)
+		require.NoError(t, err)
+		assert.Equal(t, created.ID, got.ID)
+		assert.Equal(t, "work", got.Name)
+	})
+}
+
+func TestServiceProfile_IDPrefix(t *testing.T) {
+	withTestSM(t, func(sm *ServiceManager, username string) {
+		created, err := sm.AddProfile("work", username)
+		require.NoError(t, err)
+
+		prefix := created.ID[:4]
+		got, err := sm.ResolveProfile(prefix.String(), username)
+		require.NoError(t, err)
+		assert.Equal(t, created.ID, got.ID)
+	})
+}
+
+func TestServiceProfile_AmbiguousPrefix(t *testing.T) {
+	withTestSM(t, func(sm *ServiceManager, username string) {
+		// Plant two profiles whose IDs share a known prefix by writing
+		// the files directly, since generated IDs are random.
+		configDir, err := sm.getConfigDir(username)
+		require.NoError(t, err)
+		for _, id := range []string{"abcd1111aaaa", "abcd2222bbbb"} {
+			path := filepath.Join(configDir, id+".json")
+			require.NoError(t, util.WriteJson(context.Background(), path, &Config{Name: id}))
+		}
+
+		_, err = sm.ResolveProfile("abcd", username)
+		var amb *ErrAmbiguousHandle
+		require.ErrorAs(t, err, &amb)
+		assert.Equal(t, AmbiguityKindIDPrefix, amb.Kind)
+		assert.Len(t, amb.Candidates, 2)
+	})
+}
+
+func TestServiceProfile_ExactNameUnique(t *testing.T) {
+	withTestSM(t, func(sm *ServiceManager, username string) {
+		_, err := sm.AddProfile("work", username)
+		require.NoError(t, err)
+
+		got, err := sm.ResolveProfile("work", username)
+		require.NoError(t, err)
+		assert.Equal(t, "work", got.Name)
+	})
+}
+
+func TestServiceProfile_AmbiguousName(t *testing.T) {
+	withTestSM(t, func(sm *ServiceManager, username string) {
+		_, err := sm.AddProfile("work", username)
+		require.NoError(t, err)
+		_, err = sm.AddProfile("work", username)
+		require.NoError(t, err)
+
+		_, err = sm.ResolveProfile("work", username)
+		var amb *ErrAmbiguousHandle
+		require.ErrorAs(t, err, &amb)
+		assert.Equal(t, AmbiguityKindName, amb.Kind)
+		assert.Len(t, amb.Candidates, 2)
+	})
+}
+
+func TestServiceProfile_NotFound(t *testing.T) {
+	withTestSM(t, func(sm *ServiceManager, username string) {
+		_, err := sm.ResolveProfile("nope", username)
+		assert.ErrorIs(t, err, ErrProfileNotFound)
+	})
+}
+
+func TestServiceProfile_DefaultByExactID(t *testing.T) {
+	withTestSM(t, func(sm *ServiceManager, username string) {
+		got, err := sm.ResolveProfile(defaultProfileName, username)
+		require.NoError(t, err)
+		assert.Equal(t, defaultProfileName, got.ID.String())
+	})
+}
+
+func TestServiceProfile_LegacyFilenameCoexists(t *testing.T) {
+	// Legacy profiles stored as <name>.json with no "name" JSON field
+	// should still be discoverable by name and removable by name.
+	withTestSM(t, func(sm *ServiceManager, username string) {
+		configDir, err := sm.getConfigDir(username)
+		require.NoError(t, err)
+		path := filepath.Join(configDir, "legacy.json")
+		require.NoError(t, util.WriteJson(context.Background(), path, &Config{}))
+
+		got, err := sm.ResolveProfile("legacy", username)
+		require.NoError(t, err)
+		assert.Equal(t, "legacy", got.ID.String())
+		// Name falls back to the filename stem when JSON omits it.
+		assert.Equal(t, "legacy", got.Name)
+	})
+}
+
+func TestAddProfile_AllowsDuplicateWithFlag(t *testing.T) {
+	withTestSM(t, func(sm *ServiceManager, username string) {
+		first, err := sm.AddProfile("work", username)
+		require.NoError(t, err)
+
+		second, err := sm.AddProfile("work", username)
+		require.NoError(t, err)
+		assert.NotEqual(t, first.ID, second.ID)
+		assert.Equal(t, "work", second.Name)
+	})
+}
+
+func TestAddProfile_RejectsInvalidNames(t *testing.T) {
+	withTestSM(t, func(sm *ServiceManager, username string) {
+		cases := []string{
+			"",                                       // empty
+			"\x00\x01",                               // only control chars (becomes empty)
+			strings.Repeat("a", maxProfileNameLen+1), // too long
+		}
+		for _, name := range cases {
+			_, err := sm.AddProfile(name, username)
+			assert.Error(t, err, "expected error for %q", name)
+		}
+	})
+}
+
+func TestRemoveProfile_RejectsInvalidID(t *testing.T) {
+	withTestSM(t, func(sm *ServiceManager, username string) {
+		err := sm.RemoveProfile("../escape", username)
+		assert.Error(t, err)
+	})
+}
+
+func TestSanitizeDisplayName(t *testing.T) {
+	cases := []struct {
+		in      string
+		want    string
+		wantErr bool
+	}{
+		{"work", "work", false},
+		{"My Work Account", "My Work Account", false},
+		{"emoji 🚀 ok", "emoji 🚀 ok", false},
+		{"漢字テスト", "漢字テスト", false},
+		{"with\x00null", "withnull", false},
+		{"\x01\x02\x03", "", true},
+		{"", "", true},
+	}
+	for _, tc := range cases {
+		got, err := sanitizeDisplayName(tc.in)
+		if tc.wantErr {
+			assert.Error(t, err, "case %q", tc.in)
+			continue
+		}
+		assert.NoError(t, err, "case %q", tc.in)
+		assert.Equal(t, tc.want, got, "case %q", tc.in)
+	}
+}
+
+func TestIsValidProfileFilenameStem(t *testing.T) {
+	cases := []struct {
+		in   string
+		want bool
+	}{
+		{"default", true},
+		{"abc123def456", true},
+		{"legacy-name", true},
+		{"legacy_name", true},
+		{"", false},
+		{"..", false},
+		{"../etc", false},
+		{"foo/bar", false},
+		{`foo\bar`, false},
+		{"with space", false},
+		{"with.dot", false},
+		{strings.Repeat("a", maxProfileIDLen+1), false},
+	}
+	for _, tc := range cases {
+		got := IsValidProfileFilenameStem(ID(tc.in))
+		assert.Equal(t, tc.want, got, "case %q", tc.in)
+	}
+}
+
+func TestRemoveProfile_DeletesStateFile(t *testing.T) {
+	withTestSM(t, func(sm *ServiceManager, username string) {
+		created, err := sm.AddProfile("work", username)
+		require.NoError(t, err)
+
+		configDir, err := sm.getConfigDir(username)
+		require.NoError(t, err)
+		statePath := filepath.Join(configDir, created.ID.String()+".state.json")
+		require.NoError(t, os.WriteFile(statePath, []byte(`{"email":"a@b"}`), 0600))
+
+		require.NoError(t, sm.RemoveProfile(created.ID, username))
+		_, err = os.Stat(statePath)
+		assert.True(t, errors.Is(err, os.ErrNotExist), "state file should be removed")
+	})
+}

client/internal/profilemanager/state.go

@@ -13,13 +13,20 @@ type ProfileState struct {
 	Email string `json:"email"`
 }
 
-func (pm *ProfileManager) GetProfileState(profileName string) (*ProfileState, error) {
+// GetProfileState reads the per-profile state file keyed by profile ID.
+// The state file lives in the user's config directory. Legacy state files
+// keyed by the old profile name remain readable.
+func (pm *ProfileManager) GetProfileState(id ID) (*ProfileState, error) {
 	configDir, err := getConfigDir()
 	if err != nil {
 		return nil, fmt.Errorf("get config directory: %w", err)
 	}
 
-	stateFile := filepath.Join(configDir, profileName+".state.json")
+	if id != defaultProfileName && !IsValidProfileFilenameStem(id) {
+		return nil, fmt.Errorf("invalid profile ID: %q", id)
+	}
+
+	stateFile := filepath.Join(configDir, id.String()+".state.json")
 	stateFileExists, err := fileExists(stateFile)
 	if err != nil {
 		return nil, fmt.Errorf("failed to check if profile state file exists: %w", err)
@@ -51,7 +58,12 @@ func (pm *ProfileManager) SetActiveProfileState(state *ProfileState) error {
 		return fmt.Errorf("get active profile: %w", err)
 	}
 
-	stateFile := filepath.Join(configDir, activeProf.Name+".state.json")
+	id := activeProf.ID
+	if id != defaultProfileName && !IsValidProfileFilenameStem(id) {
+		return fmt.Errorf("invalid active profile ID: %q", id)
+	}
+
+	stateFile := filepath.Join(configDir, id.String()+".state.json")
 	err = util.WriteJsonWithRestrictedPermission(context.Background(), stateFile, state)
 	if err != nil {
 		return fmt.Errorf("write profile state: %w", err)

client/internal/routemanager/dnsinterceptor/handler.go

@@ -251,6 +251,14 @@ func (d *DnsInterceptor) ServeDNS(w dns.ResponseWriter, r *dns.Msg) {
 		r.MsgHdr.AuthenticatedData = true
 	}
 
+	// Advertise EDNS0 to the forwarder so it may return an Extended DNS Error
+	// describing why a lookup failed. The OPT is stripped from the reply when
+	// the original client did not request EDNS0.
+	hadEdns := r.IsEdns0() != nil
+	if !hadEdns {
+		r.SetEdns0(dns.DefaultMsgSize, false)
+	}
+
 	upstream := net.JoinHostPort(upstreamIP.String(), strconv.FormatUint(uint64(d.forwarderPort.Load()), 10))
 	ctx, cancel := context.WithTimeout(context.Background(), dnsTimeout)
 	defer cancel()
@@ -260,6 +268,13 @@ func (d *DnsInterceptor) ServeDNS(w dns.ResponseWriter, r *dns.Msg) {
 		return
 	}
 
+	if ede, ok := resutil.ExtractEDE(reply); ok {
+		resutil.SetMeta(w, "ede", fmt.Sprintf("%d %s", ede.InfoCode, ede.ExtraText))
+	}
+	if !hadEdns {
+		resutil.StripOPT(reply)
+	}
+
 	resutil.SetMeta(w, "peer", peerKey)
 
 	reply.Id = r.Id

client/internal/routemanager/manager.go

@@ -333,6 +333,8 @@ func (m *DefaultManager) Stop(stateManager *statemanager.Manager) {
 		}
 	}
 
+	m.notifier.Close()
+
 	m.mux.Lock()
 	defer m.mux.Unlock()
 	m.clientRoutes = nil

client/internal/routemanager/notifier/notifier_android.go

@@ -16,7 +16,7 @@ import (
 type Notifier struct {
 	initialRoutes []*route.Route
 	currentRoutes []*route.Route
-	fakeIPRoutes []*route.Route
+	fakeIPRoutes  []*route.Route
 
 	listener    listener.NetworkChangeListener
 	listenerMux sync.Mutex
@@ -119,3 +119,7 @@ func (n *Notifier) GetInitialRouteRanges() []string {
 	sort.Strings(initialStrings)
 	return initialStrings
 }
+
+func (n *Notifier) Close() {
+	// unused
+}

client/internal/routemanager/notifier/notifier_ios.go

@@ -3,6 +3,7 @@
 package notifier
 
 import (
+	"container/list"
 	"net/netip"
 	"slices"
 	"sort"
@@ -14,19 +15,26 @@ import (
 )
 
 type Notifier struct {
+	mu              sync.Mutex
+	cond            *sync.Cond
 	currentPrefixes []string
-
-	listener    listener.NetworkChangeListener
-	listenerMux sync.Mutex
+	listener        listener.NetworkChangeListener
+	queue           *list.List
+	closed          bool
 }
 
 func NewNotifier() *Notifier {
-	return &Notifier{}
+	n := &Notifier{
+		queue: list.New(),
+	}
+	n.cond = sync.NewCond(&n.mu)
+	go n.deliverLoop()
+	return n
 }
 
 func (n *Notifier) SetListener(listener listener.NetworkChangeListener) {
-	n.listenerMux.Lock()
-	defer n.listenerMux.Unlock()
+	n.mu.Lock()
+	defer n.mu.Unlock()
 	n.listener = listener
 }
 
@@ -43,32 +51,52 @@ func (n *Notifier) OnNewRoutes(route.HAMap) {
 }
 
 func (n *Notifier) OnNewPrefixes(prefixes []netip.Prefix) {
-	newNets := make([]string, 0)
+	newNets := make([]string, 0, len(prefixes))
 	for _, prefix := range prefixes {
 		newNets = append(newNets, prefix.String())
 	}
 
 	sort.Strings(newNets)
 
+	n.mu.Lock()
 	if slices.Equal(n.currentPrefixes, newNets) {
+		n.mu.Unlock()
 		return
 	}
-
 	n.currentPrefixes = newNets
-	n.notify()
+	routes := strings.Join(n.currentPrefixes, ",")
+	n.queue.PushBack(routes)
+	n.cond.Signal()
+	n.mu.Unlock()
 }
-func (n *Notifier) notify() {
-	n.listenerMux.Lock()
-	defer n.listenerMux.Unlock()
-	if n.listener == nil {
-		return
-	}
 
-	go func(l listener.NetworkChangeListener) {
-		l.OnNetworkChanged(strings.Join(n.currentPrefixes, ","))
-	}(n.listener)
+func (n *Notifier) Close() {
+	n.mu.Lock()
+	n.closed = true
+	n.cond.Signal()
+	n.mu.Unlock()
 }
 
 func (n *Notifier) GetInitialRouteRanges() []string {
 	return nil
 }
+
+func (n *Notifier) deliverLoop() {
+	for {
+		n.mu.Lock()
+		for n.queue.Len() == 0 && !n.closed {
+			n.cond.Wait()
+		}
+		if n.closed && n.queue.Len() == 0 {
+			n.mu.Unlock()
+			return
+		}
+		routes := n.queue.Remove(n.queue.Front()).(string)
+		l := n.listener
+		n.mu.Unlock()
+
+		if l != nil {
+			l.OnNetworkChanged(routes)
+		}
+	}
+}

client/internal/routemanager/notifier/notifier_other.go

@@ -38,3 +38,7 @@ func (n *Notifier) OnNewPrefixes(prefixes []netip.Prefix) {
 func (n *Notifier) GetInitialRouteRanges() []string {
 	return []string{}
 }
+
+func (n *Notifier) Close() {
+	// unused
+}