Description

When a config change triggers a gateway reload, the Discord gateway's onAbort handler sets maxAttempts: 0 and disconnects. The resulting Max reconnect attempts (0) reached after code 1005 error is not caught and propagates to the global uncaughtException handler, which calls process.exit(1). This causes the gateway to crash and restart on every config reload.

Reproduction

1. Run the gateway as a systemd service with Discord integration enabled
2. Trigger any config change that requires a gateway restart (e.g., modifying auth.profiles, plugins, gateway.tools, etc.)
3. The gateway crashes with:

   [openclaw] Uncaught exception: Error: Max reconnect attempts (0) reached after code 1005

4. systemd restarts the service, and any subsequent config change repeats the cycle

Root Cause

In provider-CAlWEl41.js, the onAbort handler (around line 6952) intentionally sets gateway.options.reconnect = { maxAttempts: 0 } before calling gateway.disconnect(). This is correct behavior for a clean shutdown.

However, the error emitted by handleReconnectionAttempt when maxAttempts is reached:

this.emitter.emit("error", new Error(`Max reconnect attempts (${maxAttempts}) reached...`));

...is classified by classifyDiscordGatewayEvent as "reconnect-exhausted" with shouldStopLifecycle: true. This error is not caught by the abort flow and bubbles up as an uncaught exception, hitting:

process.on("uncaughtException", (error) => {
    console.error("[openclaw] Uncaught exception:", formatUncaughtError(error));
    process.exit(1);
});

Expected Behavior

When the abort signal fires and the gateway disconnects intentionally, the reconnect-exhausted error should be suppressed or caught, since the disconnect was deliberate. The process should not crash.

Suggested Fix

Either:

1. In the onAbort handler, set a flag (e.g., lifecycleStopping = true is already set) and check it before emitting the reconnect-exhausted error
2. Catch the error emitted during abort-triggered disconnect so it doesn't propagate to the global handler
3. Skip the handleReconnectionAttempt path entirely when maxAttempts is 0 (treat it as an intentional no-reconnect)

Environment

• openclaw v2026.3.24
• Linux (systemd user service)
• Discord gateway integration enabled
• Multiple agents configured

Impact

In the last 24 hours, this caused 83 restarts of the gateway service. Each restart drops all active Discord/Telegram sessions, interrupts embedded agent runs, and triggers re-discovery of all provider models.

Logs

Mar 26 23:34:18 [reload] config change requires gateway restart (auth.profiles.modelstudio:default)
Mar 26 23:34:18 [openclaw] Uncaught exception: Error: Max reconnect attempts (0) reached after code 1005
Mar 26 23:34:18 openclaw-gateway.service: Main process exited, code=exited, status=1/FAILURE
Mar 26 23:34:23 openclaw-gateway.service: Scheduled restart job, restart counter is at 1.
Mar 26 23:34:23 Started openclaw-gateway.service
Mar 26 23:35:00 [openclaw] Uncaught exception: Error: Max reconnect attempts (0) reached after code 1005
Mar 26 23:35:00 openclaw-gateway.service: Main process exited, code=exited, status=1/FAILURE