PR #65758: fix(whatsapp): add reconnect safety timer to connection controller

• Repository: openclaw/openclaw
• Author: nxmxbbd
• State: open | merged: False
• Link: https://github.com/openclaw/openclaw/pull/65758

Description (problem / solution / changelog)

Problem

When the WhatsApp watchdog triggers a reconnect, closeCurrentConnection() nulls the active listener immediately while the reconnect loop sleeps before re-registering. During this gap (up to 30s+ with exponential backoff), inbound messages are silently dropped because there is no active listener to handle them.

There is no mechanism to detect when this gap exceeds a reasonable threshold, so stuck reconnects go unnoticed.

Fix

Add a reconnect safety timer to WhatsAppConnectionController that detects when the listener gap exceeds 90 seconds:

• closeCurrentConnection() starts a 90s safety timer
• openConnection() clears the timer on successful reconnect
• shutdown() clears the timer to prevent dangling timers after exit
• If the timer fires, it logs a warning indicating the reconnect may be stuck

The timer is scoped to the controller instance and properly cleaned up in all code paths.

Tests

Three new tests in connection-controller.test.ts:

• Timer fires after 90s timeout when reconnect doesn't complete
• Timer is cleared when a new connection opens before timeout
• Timer is cleared on controller shutdown

Related

This PR addresses one root cause behind the widely reported "No active WhatsApp Web listener" family of issues — specifically the reconnect gap where the listener is null while the watchdog reconnect loop is in progress.

See also #65759 (companion PR — adds application-level health probe to detect silent connection failures).

Related user reports:

• #61250, #56363, #56089 — outbound fails despite connected status
• #52768, #50231, #50208 — message tool fails with "No active WhatsApp Web listener"
• #53162, #51493 — cron/announce delivery fails during reconnect
• #50383, #50489 — proactive sends silently fail
• #38734 (closed) — original report, partially addressed by #54232 (shared-map fix)

Supersedes #62000 (adapted to current controller-based architecture).

Changed files

• extensions/whatsapp/src/connection-controller.test.ts (modified, +100/-0)
• extensions/whatsapp/src/connection-controller.ts (modified, +24/-0)

---

PR #65759: fix(whatsapp): add application-level health probe to inbound monitor

• Repository: openclaw/openclaw
• Author: nxmxbbd
• State: open | merged: False
• Link: https://github.com/openclaw/openclaw/pull/65759

Description (problem / solution / changelog)

Problem

The WhatsApp inbound monitor relies on the Baileys socket connection state, but a connected socket doesn't guarantee the application layer is healthy. The socket can remain open while the server silently stops delivering messages (e.g. after a server-side session migration or network partition recovery).

The existing watchdog only checks message recency — if no messages arrive in a quiet chat, it can't distinguish "no one is messaging" from "messages aren't being delivered."

Fix

Add a periodic application-level health probe that calls sock.fetchStatus(selfJid) to verify end-to-end delivery capability:

• Probe runs on each watchdog check interval (default 60s)
• Single-flight guard prevents concurrent probes from piling up
• 15s timeout per probe — if fetchStatus hangs, the probe is marked failed but the hung call is left to settle on its own (no leaked promises)
• 3 consecutive failures trigger a watchdog-style forced reconnect
• Probe is disabled when fetchStatus is unavailable (older Baileys versions)
• Timer cleanup on connection close prevents dangling timers

Tests

monitor-health-probe.test.ts (130 lines):

• Successful probe resets failure counter
• Consecutive failures trigger forced reconnect
• Timeout handling (hung fetchStatus)
• Single-flight guard (probe already in progress)
• Timer cleanup on connection close
• Graceful degradation when fetchStatus is missing

Related

This PR addresses another root cause behind the "No active WhatsApp Web listener" family of issues — specifically silent connection failures where the socket stays alive but the application layer is dead.

See also #65758 (companion PR — adds reconnect safety timer to detect stuck reconnects).

Related user reports:

• #61250, #56363, #56089 — outbound fails despite connected status
• #52768, #50231, #50208 — message tool fails with "No active WhatsApp Web listener"
• #53162, #51493 — cron/announce delivery fails
• #50383, #50489 — proactive sends silently fail
• #38734 (closed) — original report, partially addressed by #54232 (shared-map fix)

Supersedes #61999 (clean squash onto latest main).

Changed files

• extensions/whatsapp/src/auto-reply.test-harness.ts (modified, +2/-0)
• extensions/whatsapp/src/inbound/monitor-health-probe.test.ts (added, +130/-0)
• extensions/whatsapp/src/inbound/monitor.ts (modified, +94/-0)
• extensions/whatsapp/src/monitor-inbox.test-harness.ts (modified, +1/-0)