feat(gateway): make health monitor stale threshold and max restarts configurable

src/config/schema.help.ts

@@ -102,6 +102,10 @@ export const FIELD_HELP: Record<string, string> = {
     "Explicit gateway-level tool denylist to block risky tools even if lower-level policies allow them. Use deny rules for emergency response and defense-in-depth hardening.",
   "gateway.channelHealthCheckMinutes":
     "Interval in minutes for automatic channel health probing and status updates. Use lower intervals for faster detection, or higher intervals to reduce periodic probe noise.",
+  "gateway.channelStaleEventThresholdMinutes":
+    "How many minutes a connected channel can go without receiving any event before the health monitor treats it as a stale socket and triggers a restart. Default: 30.",
+  "gateway.channelMaxRestartsPerHour":
+    "Maximum number of health-monitor-initiated channel restarts allowed within a rolling one-hour window. Once hit, further restarts are skipped until the window expires. Default: 10.",
   "gateway.tailscale":
     "Tailscale integration settings for Serve/Funnel exposure and lifecycle handling on gateway start/exit. Keep off unless your deployment intentionally relies on Tailscale ingress.",
   "gateway.tailscale.mode":

src/config/schema.labels.ts

@@ -84,6 +84,8 @@ export const FIELD_LABELS: Record<string, string> = {
   "gateway.tools.allow": "Gateway Tool Allowlist",
   "gateway.tools.deny": "Gateway Tool Denylist",
   "gateway.channelHealthCheckMinutes": "Gateway Channel Health Check Interval (min)",
+  "gateway.channelStaleEventThresholdMinutes": "Gateway Channel Stale Event Threshold (min)",
+  "gateway.channelMaxRestartsPerHour": "Gateway Channel Max Restarts Per Hour",
   "gateway.tailscale": "Gateway Tailscale",
   "gateway.tailscale.mode": "Gateway Tailscale Mode",
   "gateway.tailscale.resetOnExit": "Gateway Tailscale Reset on Exit",

src/config/types.gateway.ts

@@ -431,4 +431,16 @@ export type GatewayConfig = {
    * Set to 0 to disable. Default: 5.
    */
   channelHealthCheckMinutes?: number;
+  /**
+   * Stale event threshold in minutes for the channel health monitor.
+   * A connected channel that receives no events for this duration is treated
+   * as a stale socket and restarted. Default: 30.
+   */
+  channelStaleEventThresholdMinutes?: number;
+  /**
+   * Maximum number of health-monitor-initiated channel restarts per hour.
+   * Once this limit is reached, the monitor skips further restarts until
+   * the rolling window expires. Default: 10.
+   */
+  channelMaxRestartsPerHour?: number;
 };

src/config/zod-schema.ts

@@ -696,6 +696,8 @@ export const OpenClawSchema = z
           .strict()
           .optional(),
         channelHealthCheckMinutes: z.number().int().min(0).optional(),
+        channelStaleEventThresholdMinutes: z.number().int().min(1).optional(),
+        channelMaxRestartsPerHour: z.number().int().min(1).optional(),
         tailscale: z
           .object({
             mode: z.union([z.literal("off"), z.literal("serve"), z.literal("funnel")]).optional(),

src/gateway/config-reload-plan.ts

@@ -41,6 +41,16 @@ const BASE_RELOAD_RULES: ReloadRule[] = [
     kind: "hot",
     actions: ["restart-health-monitor"],
   },
+  {
+    prefix: "gateway.channelStaleEventThresholdMinutes",
+    kind: "hot",
+    actions: ["restart-health-monitor"],
+  },
+  {
+    prefix: "gateway.channelMaxRestartsPerHour",
+    kind: "hot",
+    actions: ["restart-health-monitor"],
+  },
   // Stuck-session warning threshold is read by the diagnostics heartbeat loop.
   { prefix: "diagnostics.stuckSessionWarnMs", kind: "none" },
   { prefix: "hooks.gmail", kind: "hot", actions: ["restart-gmail-watcher"] },

src/gateway/server-reload-handlers.ts

@@ -50,7 +50,11 @@ export function createGatewayReloadHandlers(params: {
   logChannels: { info: (msg: string) => void; error: (msg: string) => void };
   logCron: { error: (msg: string) => void };
   logReload: { info: (msg: string) => void; warn: (msg: string) => void };
-  createHealthMonitor: (checkIntervalMs: number) => ChannelHealthMonitor;
+  createHealthMonitor: (opts: {
+    checkIntervalMs: number;
+    staleEventThresholdMs?: number;
+    maxRestartsPerHour?: number;
+  }) => ChannelHealthMonitor;
 }) {
   const applyHotReload = async (
     plan: GatewayReloadPlan,
@@ -101,8 +105,17 @@ export function createGatewayReloadHandlers(params: {
     if (plan.restartHealthMonitor) {
       state.channelHealthMonitor?.stop();
       const minutes = nextConfig.gateway?.channelHealthCheckMinutes;
+      const staleMinutes = nextConfig.gateway?.channelStaleEventThresholdMinutes;
       nextState.channelHealthMonitor =
-        minutes === 0 ? null : params.createHealthMonitor((minutes ?? 5) * 60_000);
+        minutes === 0
+          ? null
+          : params.createHealthMonitor({
+              checkIntervalMs: (minutes ?? 5) * 60_000,
+              ...(staleMinutes != null && { staleEventThresholdMs: staleMinutes * 60_000 }),
+              ...(nextConfig.gateway?.channelMaxRestartsPerHour != null && {
+                maxRestartsPerHour: nextConfig.gateway.channelMaxRestartsPerHour,
+              }),
+            });
     }
 
     if (plan.restartGmailWatcher) {

src/gateway/server.impl.ts

@@ -757,11 +757,17 @@ export async function startGatewayServer(
 
   const healthCheckMinutes = cfgAtStart.gateway?.channelHealthCheckMinutes;
   const healthCheckDisabled = healthCheckMinutes === 0;
+  const staleEventThresholdMinutes = cfgAtStart.gateway?.channelStaleEventThresholdMinutes;
+  const maxRestartsPerHour = cfgAtStart.gateway?.channelMaxRestartsPerHour;
   let channelHealthMonitor = healthCheckDisabled
     ? null
     : startChannelHealthMonitor({
         channelManager,
         checkIntervalMs: (healthCheckMinutes ?? 5) * 60_000,
+        ...(staleEventThresholdMinutes != null && {
+          staleEventThresholdMs: staleEventThresholdMinutes * 60_000,
+        }),
+        ...(maxRestartsPerHour != null && { maxRestartsPerHour }),
       });
 
   if (!minimalTestGateway) {
@@ -980,8 +986,21 @@ export async function startGatewayServer(
           logChannels,
           logCron,
           logReload,
-          createHealthMonitor: (checkIntervalMs: number) =>
-            startChannelHealthMonitor({ channelManager, checkIntervalMs }),
+          createHealthMonitor: (opts: {
+            checkIntervalMs: number;
+            staleEventThresholdMs?: number;
+            maxRestartsPerHour?: number;
+          }) =>
+            startChannelHealthMonitor({
+              channelManager,
+              checkIntervalMs: opts.checkIntervalMs,
+              ...(opts.staleEventThresholdMs != null && {
+                staleEventThresholdMs: opts.staleEventThresholdMs,
+              }),
+              ...(opts.maxRestartsPerHour != null && {
+                maxRestartsPerHour: opts.maxRestartsPerHour,
+              }),
+            }),
         });
 
         return startGatewayConfigReloader({