Add Windows NRPT recovery circuit breaker

Windows DNS intercept mode runs an NRPT health monitor that restores the
catch-all rule and re-signals DNS Client whenever Windows stops routing
queries to the local listener. When another agent (MDM, VPN, GPO) keeps
putting NRPT back into a broken state, that loop never converges: ctrld
repeatedly calls RefreshPolicyEx, Dnscache paramchange, and flushes the
DNS cache, producing continuous flash writes and SIEM noise while never
fixing anything.

Add a recovery limiter that trips after a configurable number of
consecutive recovery flows and enters a cooldown, during which recovery
is suppressed (logged at most once every 5 minutes). Clearing the
circuit requires two consecutive stable health successes rather than
one, because a probe can pass briefly right after delete/re-add even
when the underlying NRPT state is still broken.

New [service] options gate the behavior and default to the previous
unlimited behavior:
  - nrpt_recovery_max_attempts (default 0 = unlimited)
  - nrpt_recovery_cooldown     (default 30m)

Also collapse the repeated refresh + paramchange + flush sequence into a
single signalNRPTChange() helper, and make cleanGPPath /
cleanEmptyNRPTParent only mutate the registry and report whether cleanup
happened, so callers send exactly one DNS Client change signal instead
of several. When the GP DnsPolicyConfig parent exists but is empty,
nrptProbeAndHeal now cleans it and signals once before spending the
normal policy-refresh retry budget, since those retries cannot succeed
while DNS Client is stuck in GP mode.

Add unit tests for the limiter's cooldown, stable-reset, and unlimited
paths, and document the new options and the empty-GP repro.

cmd/cli/dns_intercept_windows.go

@@ -278,6 +278,9 @@ type wfpState struct {
 	loopbackProtectActive bool
 	// loopbackPermitIDs stores the filter IDs for the loopback protect permits.
 	loopbackPermitIDs []uint64
+	// nrptRecoveryLimiter prevents repeated Windows policy/Dnscache signaling
+	// when another agent keeps putting NRPT back into a broken state.
+	nrptRecoveryLimiter nrptRecoveryLimiter
 }
 
 // Lazy-loaded WFP DLL procedures.
@@ -343,9 +346,10 @@ const (
 //   - GP path: SOFTWARE\Policies\...\DnsPolicyConfig (Group Policy)
 //   - Local path: SYSTEM\CurrentControlSet\...\DnsPolicyConfig (service store)
 //
-// If ANY rules exist in the GP path (from IT policy, VPN, MDM, etc.), DNS Client
+// If the GP path contains real rules (from IT policy, VPN, MDM, etc.), DNS
-// enters "GP mode" and ignores ALL local-path rules entirely. Conversely, if the
+// Client enters "GP mode" and ignores ALL local-path rules entirely. An empty GP
-// GP path is empty/absent, DNS Client reads from the local path only.
+// parent key is worse: it still puts DNS Client in GP mode, but contributes no
+// usable rule, so our local catch-all is hidden until that empty parent is gone.
 //
 // Strategy (matching Tailscale's approach):
 //   - Always write to the local path (baseline for non-domain machines).
@@ -395,18 +399,20 @@ func otherGPRulesExist() bool {
 	return false
 }
 
-// cleanGPPath removes our CtrldCatchAll rule from the GP path and deletes
+// cleanGPPath removes only ctrld's GP-path rule and deletes the GP parent when
-// the GP DnsPolicyConfig parent key if no other rules remain. Removing the
+// no rules remain. The return value tells callers whether the parent key was
-// empty GP key is critical: its mere existence forces DNS Client into "GP mode"
+// actually deleted, which means DNS Client should be signaled once.
-// where local-path rules are ignored.
+//
-func cleanGPPath() {
+// Do not leave an empty GP parent behind: Windows treats the parent key itself
+// as the policy store boundary, so an empty key can still hide local-path rules.
+func cleanGPPath() bool {
 	// Delete our specific rule.
 	registry.DeleteKey(registry.LOCAL_MACHINE, nrptBaseKey+`\`+nrptRuleName)
 
 	// If the GP parent key is now empty, delete it entirely to exit "GP mode".
 	k, err := registry.OpenKey(registry.LOCAL_MACHINE, nrptBaseKey, registry.ENUMERATE_SUB_KEYS)
 	if err != nil {
-		return // Key doesn't exist — clean state.
+		return false // Key doesn't exist — clean state.
 	}
 	names, err := k.ReadSubKeyNames(-1)
 	k.Close()
@@ -414,12 +420,14 @@ func cleanGPPath() {
 		if len(names) > 0 {
 			mainLog.Load().Debug().Strs("remaining", names).Msg("DNS intercept: GP path has other rules, leaving parent key")
 		}
-		return
+		return false
 	}
 	// Empty — delete it to exit "GP mode".
 	if err := registry.DeleteKey(registry.LOCAL_MACHINE, nrptBaseKey); err == nil {
 		mainLog.Load().Info().Msg("DNS intercept: deleted empty GP DnsPolicyConfig key (exits GP mode)")
+		return true
 	}
+	return false
 }
 
 // writeNRPTRule writes a single NRPT catch-all rule at the given registry keyPath.
@@ -542,10 +550,11 @@ func refreshNRPTPolicy() {
 // Group Policy refresh so NRPT changes take effect immediately.
 // Uses DnsFlushResolverCache from dnsapi.dll + RefreshPolicyEx from userenv.dll.
 func flushDNSCache() {
-	// Step 1: Refresh GP so DNS Client loads the new NRPT rules from registry.
 	refreshNRPTPolicy()
+	flushDNSCacheOnly()
+}
 
-	// Step 2: Flush the DNS cache so stale entries from pre-NRPT resolution are cleared.
+func flushDNSCacheOnly() {
 	if err := dnsapiDLL.Load(); err == nil {
 		if err := procDnsFlushResolverCache.Find(); err == nil {
 			ret, _, _ := procDnsFlushResolverCache.Call()
@@ -555,7 +564,6 @@ func flushDNSCache() {
 			}
 		}
 	}
-	// Fallback: use ipconfig /flushdns.
 	if out, err := exec.Command("ipconfig", "/flushdns").CombinedOutput(); err != nil {
 		mainLog.Load().Debug().Msgf("DNS intercept: ipconfig /flushdns failed: %v: %s", err, string(out))
 	} else {
@@ -563,6 +571,12 @@ func flushDNSCache() {
 	}
 }
 
+func signalNRPTChange() {
+	refreshNRPTPolicy()
+	sendParamChange()
+	flushDNSCacheOnly()
+}
+
 // startDNSIntercept activates WFP-based DNS interception on Windows.
 // It creates a WFP sublayer and adds filters that block all outbound DNS (port 53)
 // traffic except to localhost (127.0.0.1/::1), ensuring all DNS queries must go
@@ -598,21 +612,19 @@ func (p *prog) startDNSIntercept() error {
 	mainLog.Load().Info().Msgf("DNS intercept: initializing (mode: %s)", interceptMode)
 	logNRPTParentKeyState("pre-write")
 
-	// Two-phase empty parent key recovery: if the GP DnsPolicyConfig key exists
+	// Empty parent key recovery: if the GP DnsPolicyConfig key exists but is
-	// but is empty, DNS Client has cached a "no rules" state and won't accept
+	// empty, DNS Client enters GP mode and hides local rules. Delete empty
-	// new rules even after they're written. Delete the empty key and signal DNS
+	// parents first, then send one change signal so DNS Client drops stale state.
-	// Client to reset before writing our rule.
+	if cleanEmptyNRPTParent() {
-	// Two-phase recovery handles its own 2s signaling burst internally.
+		signalNRPTChange()
-	cleanEmptyNRPTParent()
+	}
 
 	if err := addNRPTCatchAllRule(listenerIP); err != nil {
 		return fmt.Errorf("dns intercept: failed to add NRPT catch-all rule: %w", err)
 	}
 	logNRPTParentKeyState("post-write")
 	state.nrptActive = true
-	refreshNRPTPolicy()
+	signalNRPTChange()
-	sendParamChange()
-	flushDNSCache()
 	mainLog.Load().Info().Msgf("DNS intercept: NRPT catch-all rule active — all DNS queries directed to %s", listenerIP)
 
 	// Step 2: In hard mode, also set up WFP filters to block non-local DNS.
@@ -1555,7 +1567,7 @@ func (p *prog) scheduleDelayedRechecks() {
 					mainLog.Load().Error().Err(err).Msg("DNS intercept: failed to re-add NRPT catch-all rule")
 					state.nrptActive = false
 				} else {
-					flushDNSCache()
+					signalNRPTChange()
 					mainLog.Load().Info().Msg("DNS intercept: NRPT catch-all rule restored")
 				}
 			}
@@ -1593,14 +1605,22 @@ func (p *prog) nrptHealthMonitor(state *wfpState) {
 
 			// Step 1: Check registry key exists.
 			if !nrptCatchAllRuleExists() {
+				now := time.Now()
+				if ok, wait := state.nrptRecoveryLimiter.allow(now, p.cfg); !ok {
+					if state.nrptRecoveryLimiter.shouldLogSkip(now) {
+						mainLog.Load().Warn().Dur("remaining", wait).
+							Msg("DNS intercept: NRPT rule restore suppressed after repeated recovery flows")
+					}
+					continue
+				}
 				mainLog.Load().Warn().Msg("DNS intercept: NRPT health check — catch-all rule missing, restoring")
 				if err := addNRPTCatchAllRule(state.listenerIP); err != nil {
 					mainLog.Load().Error().Err(err).Msg("DNS intercept: failed to restore NRPT catch-all rule")
 					state.nrptActive = false
 					continue
 				}
-				refreshNRPTPolicy()
+				signalNRPTChange()
-				flushDNSCache()
+				state.nrptRecoveryLimiter.recordRecoveryFlow(time.Now(), p.cfg)
 				mainLog.Load().Info().Msg("DNS intercept: NRPT catch-all rule restored by health monitor")
 				// After restoring, verify it's actually working.
 				go p.nrptProbeAndHeal()
@@ -1612,6 +1632,8 @@ func (p *prog) nrptHealthMonitor(state *wfpState) {
 			if !p.probeNRPT() {
 				mainLog.Load().Warn().Msg("DNS intercept: NRPT health check — rule present but probe failed, running heal cycle")
 				go p.nrptProbeAndHeal()
+			} else {
+				state.nrptRecoveryLimiter.recordStableSuccess()
 			}
 
 			// Step 3: In hard mode, also verify WFP sublayer.
@@ -1710,43 +1732,39 @@ func sendParamChange() {
 }
 
 // cleanEmptyNRPTParent removes empty NRPT parent keys that block activation.
-// An empty DnsPolicyConfig key (exists but no subkeys) causes DNS Client to
+// Empty GP and local parents have different failure shapes:
-// cache "no rules" and ignore subsequently-added rules.
+//   - empty GP parent: DNS Client is in GP mode and ignores local-path rules;
+//   - empty local parent: DNS Client can cache an empty local policy store.
 //
-// Also cleans the GP path entirely if it has no non-ctrld rules, since the GP
+// This helper only changes registry state. The caller sends the single
-// path's existence forces DNS Client into "GP mode" where it ignores the local
+// RefreshPolicyEx/paramchange/flush signal after it knows cleanup occurred.
-// service store path.
 //
-// Returns true if cleanup was performed (caller should add a delay).
+// Returns true if cleanup was performed (caller should signal DNS Client).
 func cleanEmptyNRPTParent() bool {
-	cleaned := false
-
 	// Always clean the GP path — its existence blocks local path activation.
-	cleanGPPath()
+	cleaned := cleanGPPath()
 
 	// Clean empty local/direct path parent key.
-	k, err := registry.OpenKey(registry.LOCAL_MACHINE, nrptDirectKey, registry.ENUMERATE_SUB_KEYS)
+	if !nrptParentKeyEmpty(nrptDirectKey) {
-	if err != nil {
+		return cleaned
-		return false
-	}
-	names, err := k.ReadSubKeyNames(-1)
-	k.Close()
-	if err != nil || len(names) > 0 {
-		return false
 	}
 
 	mainLog.Load().Warn().Msg("DNS intercept: found empty NRPT local parent key (blocks activation) — removing")
 	if err := registry.DeleteKey(registry.LOCAL_MACHINE, nrptDirectKey); err != nil {
 		mainLog.Load().Warn().Err(err).Msg("DNS intercept: failed to delete empty NRPT local parent key")
+		return cleaned
+	}
+	return true
+}
+
+func nrptParentKeyEmpty(keyPath string) bool {
+	k, err := registry.OpenKey(registry.LOCAL_MACHINE, keyPath, registry.ENUMERATE_SUB_KEYS)
+	if err != nil {
 		return false
 	}
-	cleaned = true
+	names, err := k.ReadSubKeyNames(-1)
-
+	k.Close()
-	// Signal DNS Client to process the deletion and reset its internal cache.
+	return err == nil && len(names) == 0
-	mainLog.Load().Info().Msg("DNS intercept: empty NRPT parent key removed — signaling DNS Client")
-	sendParamChange()
-	flushDNSCache()
-	return cleaned
 }
 
 // logNRPTParentKeyState logs the state of both NRPT registry paths for diagnostics.
@@ -1783,18 +1801,39 @@ func logNRPTParentKeyState(context string) {
 // nrptProbeAndHeal runs the NRPT probe with retries and escalating remediation.
 // Called asynchronously after startup and from the health monitor.
 //
-// Retry sequence (each attempt: GP refresh + paramchange + flush → sleep → probe):
+// Retry sequence:
-//  1. Immediate probe
+//  1. Immediate probe.
-//  2. GP refresh + paramchange + flush → 1s → probe
+//  2. If the GP parent is empty, clean it immediately, signal once, then probe.
-//  3. GP refresh + paramchange + flush → 2s → probe
+//     This is intentionally before the normal retry loop: policy refresh and
-//  4. GP refresh + paramchange + flush → 4s → probe
+//     Dnscache paramchange cannot make local rules visible while GP mode is
+//     selected by an empty GP parent.
+//  3. Otherwise, signal DNS Client with increasing backoff between probes.
 func (p *prog) nrptProbeAndHeal() {
+	state, _ := p.dnsInterceptState.(*wfpState)
+	if state != nil {
+		now := time.Now()
+		if ok, wait := state.nrptRecoveryLimiter.allow(now, p.cfg); !ok {
+			if state.nrptRecoveryLimiter.shouldLogSkip(now) {
+				mainLog.Load().Warn().Dur("remaining", wait).
+					Msg("DNS intercept: NRPT recovery suppressed after repeated failed recovery flows")
+			}
+			return
+		}
+	}
+
 	if !nrptProbeRunning.CompareAndSwap(false, true) {
 		mainLog.Load().Debug().Msg("DNS intercept: NRPT probe already running, skipping")
 		return
 	}
 	defer nrptProbeRunning.Store(false)
 
+	remediated := false
+	defer func() {
+		if remediated && state != nil {
+			state.nrptRecoveryLimiter.recordRecoveryFlow(time.Now(), p.cfg)
+		}
+	}()
+
 	mainLog.Load().Info().Msg("DNS intercept: starting NRPT verification probe sequence")
 
 	// Log parent key state for diagnostics.
@@ -1805,17 +1844,37 @@ func (p *prog) nrptProbeAndHeal() {
 		mainLog.Load().Info().Msg("DNS intercept: NRPT verified working")
 		return
 	}
+	remediated = true
 
-	// Attempts 2-4: GP refresh + paramchange + flush with increasing backoff
+	// If the GP parent exists but is empty, do not burn retries on Windows
+	// signaling. Those retries create SIEM noise but cannot succeed because DNS
+	// Client is still reading the empty GP store instead of the populated local
+	// store. Delete the blocker, send one notification, then re-probe.
+	if nrptParentKeyEmpty(nrptBaseKey) {
+		mainLog.Load().Warn().Msg("DNS intercept: NRPT probe failed with empty GP parent — cleaning before retry signaling")
+		if cleanEmptyNRPTParent() {
+			signalNRPTChange()
+			time.Sleep(1 * time.Second)
+			logNRPTParentKeyState("empty-gp-after-clean")
+			if p.probeNRPT() {
+				mainLog.Load().Info().Msg("DNS intercept: NRPT verified working after empty GP parent cleanup")
+				return
+			}
+		}
+		if nrptParentKeyEmpty(nrptBaseKey) {
+			mainLog.Load().Warn().Msg("DNS intercept: empty GP NRPT parent still present after cleanup; skipping redundant policy refresh retries")
+			return
+		}
+	}
+
+	// Attempts 2-4: signal DNS Client with increasing backoff between probes.
 	delays := []time.Duration{1 * time.Second, 2 * time.Second, 4 * time.Second}
 	for i, delay := range delays {
 		attempt := i + 2
 		mainLog.Load().Info().Int("attempt", attempt).Dur("delay", delay).
-			Msg("DNS intercept: NRPT probe failed, retrying with GP refresh + paramchange")
+			Msg("DNS intercept: NRPT probe failed, retrying with policy refresh + paramchange")
 		logNRPTParentKeyState(fmt.Sprintf("probe-attempt-%d", attempt))
-		refreshNRPTPolicy()
+		signalNRPTChange()
-		sendParamChange()
-		flushDNSCache()
 		time.Sleep(delay)
 		if p.probeNRPT() {
 			mainLog.Load().Info().Int("attempt", attempt).
@@ -1829,7 +1888,7 @@ func (p *prog) nrptProbeAndHeal() {
 	// signal DNS Client to forget it, wait, then re-add and signal again.
 	mainLog.Load().Warn().Msg("DNS intercept: all probes failed — attempting two-phase NRPT recovery (delete → signal → re-add)")
 	listenerIP := "127.0.0.1"
-	if state, ok := p.dnsInterceptState.(*wfpState); ok {
+	if state != nil {
 		listenerIP = state.listenerIP
 	}
 
@@ -1837,9 +1896,7 @@ func (p *prog) nrptProbeAndHeal() {
 	_ = removeNRPTCatchAllRule()
 	// If parent key is now empty after removing our rule, delete it too.
 	cleanEmptyNRPTParent()
-	refreshNRPTPolicy()
+	signalNRPTChange()
-	sendParamChange()
-	flushDNSCache()
 	logNRPTParentKeyState("nuclear-after-delete")
 
 	// Wait for DNS Client to process the deletion.
@@ -1850,9 +1907,7 @@ func (p *prog) nrptProbeAndHeal() {
 		mainLog.Load().Error().Err(err).Msg("DNS intercept: failed to re-add NRPT after nuclear recovery")
 		return
 	}
-	refreshNRPTPolicy()
+	signalNRPTChange()
-	sendParamChange()
-	flushDNSCache()
 	logNRPTParentKeyState("nuclear-after-readd")
 
 	// Final probe after recovery.

cmd/cli/nrpt_recovery_limiter_windows.go

@@ -0,0 +1,101 @@
+//go:build windows
+
+package cli
+
+import (
+	"sync"
+	"time"
+
+	"github.com/Control-D-Inc/ctrld"
+)
+
+const (
+	// Default to current behavior: keep recovering indefinitely unless configured.
+	defaultNRPTRecoveryMaxAttempts = 0
+	defaultNRPTRecoveryCooldown    = 30 * time.Minute
+
+	// Require more than one good health tick before clearing the circuit. A probe can
+	// pass briefly after delete/re-add even when another agent recreates broken NRPT state.
+	nrptRecoveryStableSuccessesToReset = 2
+)
+
+type nrptRecoveryLimiter struct {
+	mu              sync.Mutex
+	attempts        int
+	stableSuccesses int
+	cooldownUntil   time.Time
+	lastSkipLog     time.Time
+}
+
+func nrptRecoveryMaxAttempts(cfg *ctrld.Config) int {
+	if cfg != nil && cfg.Service.NRPTRecoveryMaxAttempts != nil {
+		return *cfg.Service.NRPTRecoveryMaxAttempts
+	}
+	return defaultNRPTRecoveryMaxAttempts
+}
+
+func nrptRecoveryCooldown(cfg *ctrld.Config) time.Duration {
+	if cfg != nil && cfg.Service.NRPTRecoveryCooldown != nil {
+		return *cfg.Service.NRPTRecoveryCooldown
+	}
+	return defaultNRPTRecoveryCooldown
+}
+
+func (l *nrptRecoveryLimiter) allow(now time.Time, cfg *ctrld.Config) (bool, time.Duration) {
+	maxAttempts := nrptRecoveryMaxAttempts(cfg)
+	if maxAttempts <= 0 {
+		return true, 0
+	}
+
+	l.mu.Lock()
+	defer l.mu.Unlock()
+
+	if now.Before(l.cooldownUntil) {
+		return false, l.cooldownUntil.Sub(now)
+	}
+	return true, 0
+}
+
+func (l *nrptRecoveryLimiter) recordRecoveryFlow(now time.Time, cfg *ctrld.Config) {
+	maxAttempts := nrptRecoveryMaxAttempts(cfg)
+	if maxAttempts <= 0 {
+		return
+	}
+
+	cooldown := nrptRecoveryCooldown(cfg)
+	if cooldown <= 0 {
+		cooldown = defaultNRPTRecoveryCooldown
+	}
+
+	l.mu.Lock()
+	defer l.mu.Unlock()
+
+	l.stableSuccesses = 0
+	l.attempts++
+	if l.attempts >= maxAttempts {
+		l.cooldownUntil = now.Add(cooldown)
+	}
+}
+
+func (l *nrptRecoveryLimiter) recordStableSuccess() {
+	l.mu.Lock()
+	defer l.mu.Unlock()
+
+	l.stableSuccesses++
+	if l.stableSuccesses >= nrptRecoveryStableSuccessesToReset {
+		l.attempts = 0
+		l.cooldownUntil = time.Time{}
+		l.lastSkipLog = time.Time{}
+	}
+}
+
+func (l *nrptRecoveryLimiter) shouldLogSkip(now time.Time) bool {
+	l.mu.Lock()
+	defer l.mu.Unlock()
+
+	if l.lastSkipLog.IsZero() || now.Sub(l.lastSkipLog) >= 5*time.Minute {
+		l.lastSkipLog = now
+		return true
+	}
+	return false
+}

cmd/cli/nrpt_recovery_limiter_windows_test.go

@@ -0,0 +1,74 @@
+//go:build windows
+
+package cli
+
+import (
+	"testing"
+	"time"
+
+	"github.com/Control-D-Inc/ctrld"
+)
+
+func TestNRPTRecoveryLimiterCooldownAndStableReset(t *testing.T) {
+	maxAttempts := 2
+	cooldown := 10 * time.Minute
+	cfg := &ctrld.Config{}
+	cfg.Service.NRPTRecoveryMaxAttempts = &maxAttempts
+	cfg.Service.NRPTRecoveryCooldown = &cooldown
+
+	limiter := &nrptRecoveryLimiter{}
+	now := time.Unix(100, 0)
+
+	if ok, wait := limiter.allow(now, cfg); !ok || wait != 0 {
+		t.Fatalf("initial allow = %v, %v; want true, 0", ok, wait)
+	}
+
+	limiter.recordRecoveryFlow(now, cfg)
+	if ok, wait := limiter.allow(now.Add(time.Second), cfg); !ok || wait != 0 {
+		t.Fatalf("allow after first flow = %v, %v; want true, 0", ok, wait)
+	}
+
+	limiter.recordRecoveryFlow(now.Add(2*time.Second), cfg)
+	if ok, wait := limiter.allow(now.Add(3*time.Second), cfg); ok || wait <= 0 {
+		t.Fatalf("allow after max flows = %v, %v; want false, positive wait", ok, wait)
+	}
+
+	limiter.recordStableSuccess()
+	if ok, _ := limiter.allow(now.Add(4*time.Second), cfg); ok {
+		t.Fatal("one stable success cleared cooldown; want cooldown to remain")
+	}
+
+	limiter.recordStableSuccess()
+	if ok, wait := limiter.allow(now.Add(5*time.Second), cfg); !ok || wait != 0 {
+		t.Fatalf("allow after stable reset = %v, %v; want true, 0", ok, wait)
+	}
+}
+
+func TestNRPTRecoveryLimiterDefaultIsUnlimited(t *testing.T) {
+	cfg := &ctrld.Config{}
+	limiter := &nrptRecoveryLimiter{}
+	now := time.Unix(100, 0)
+
+	for i := 0; i < 10; i++ {
+		limiter.recordRecoveryFlow(now.Add(time.Duration(i)*time.Second), cfg)
+	}
+	if ok, wait := limiter.allow(now.Add(time.Hour), cfg); !ok || wait != 0 {
+		t.Fatalf("default allow after recovery flows = %v, %v; want true, 0", ok, wait)
+	}
+}
+
+func TestNRPTRecoveryLimiterUnlimited(t *testing.T) {
+	maxAttempts := 0
+	cfg := &ctrld.Config{}
+	cfg.Service.NRPTRecoveryMaxAttempts = &maxAttempts
+
+	limiter := &nrptRecoveryLimiter{}
+	now := time.Unix(100, 0)
+
+	for i := 0; i < 10; i++ {
+		limiter.recordRecoveryFlow(now.Add(time.Duration(i)*time.Second), cfg)
+	}
+	if ok, wait := limiter.allow(now.Add(time.Hour), cfg); !ok || wait != 0 {
+		t.Fatalf("unlimited allow = %v, %v; want true, 0", ok, wait)
+	}
+}

config.go

@@ -241,6 +241,8 @@ type ServiceConfig struct {
 	ForceRefetchWaitTime    *int           `mapstructure:"force_refetch_wait_time" toml:"force_refetch_wait_time,omitempty"`
 	LeakOnUpstreamFailure   *bool          `mapstructure:"leak_on_upstream_failure" toml:"leak_on_upstream_failure,omitempty"`
 	InterceptMode           string         `mapstructure:"intercept_mode" toml:"intercept_mode,omitempty" validate:"omitempty,oneof=off dns hard"`
+	NRPTRecoveryMaxAttempts *int           `mapstructure:"nrpt_recovery_max_attempts" toml:"nrpt_recovery_max_attempts,omitempty" validate:"omitempty,gte=0"`
+	NRPTRecoveryCooldown    *time.Duration `mapstructure:"nrpt_recovery_cooldown" toml:"nrpt_recovery_cooldown,omitempty"`
 	Daemon                  bool           `mapstructure:"-" toml:"-"`
 	AllocateIP              bool           `mapstructure:"-" toml:"-"`
 }

docs/config.md

@@ -295,6 +295,22 @@ If a remote upstream fails to resolve a query or is unreachable, `ctrld` will fo
 - Required: no
 - Default: true on Windows, MacOS and non-router Linux.
 
+### nrpt_recovery_max_attempts
+Windows DNS intercept mode uses NRPT health probes and recovery when Windows stops routing queries to the local `ctrld` listener. This limits how many consecutive recovery flows can run before `ctrld` enters a cooldown and stops making policy/Dnscache changes.
+
+Set to `0` to disable this circuit breaker and keep retrying indefinitely.
+
+- Type: integer
+- Required: no
+- Default: 0 (unlimited, current behavior)
+
+### nrpt_recovery_cooldown
+Cooldown duration after `nrpt_recovery_max_attempts` consecutive Windows NRPT recovery flows. During cooldown, `ctrld` logs the suppressed recovery and avoids additional `RefreshPolicyEx`, Dnscache `paramchange`, and DNS cache flush calls.
+
+- Type: time duration string
+- Required: no
+- Default: 30m
+
 ## Upstream
 The `[upstream]` section specifies the DNS upstream servers that `ctrld` will forward DNS requests to.
 

docs/wfp-dns-intercept.md

@@ -91,6 +91,20 @@ ctrld uses an adaptive strategy (matching [Tailscale's approach](https://github.
    the empty GP parent key. This ensures DNS Client stays in "local mode" where
    the local-path rule activates immediately via `paramchange`.
 
+### Reproducing the Empty GP Parent Case
+
+This is a production code reference, so the temporary repro script is not kept in
+the repository. For MR !942 review, the test script and exact before/after steps
+are posted in the MR discussion. The scenario to compare is:
+
+1. Run the same approved PowerShell repro script against a pre-fix build and this
+   branch with the same ctrld config.
+2. Create an empty GP NRPT parent key while ctrld is running in DNS intercept mode.
+3. Confirm pre-fix logs can spend policy refresh/paramchange retries while the GP
+   parent remains empty.
+4. Confirm post-fix logs clean the empty GP parent, send one NRPT-change signal,
+   and re-probe before normal retries.
+
 ### VPN Coexistence
 
 NRPT uses most-specific-match. VPN NRPT rules for specific domains (e.g.,