openclaw - 💡(How to fix) Fix [Bug]: ClawTalk WebSocket disconnect starves Slack Socket Mode event loop, causing stale-socket restart [1 participants]

Description

When the ClawTalk WebSocket connection (wss://clawdtalk.com/ws) experiences a pong timeout and disconnects, Slack Socket Mode message processing stops. Slack messages are not acknowledged within the 3-second window, the health monitor detects a stale socket, and restarts the Slack provider. Buffered messages may be lost.

ClawTalk and Slack are independent channels connecting to different servers, but they share the same gateway Node.js process and event loop. A transient failure in one channel should not affect the other.

Steps to Reproduce

1. Have both ClawTalk and Slack extensions enabled in openclaw.json
2. Gateway is running normally, processing Slack messages
3. ClawTalk WebSocket misses a pong (network hiccup, server maintenance, or heavy compaction event)
4. ClawTalk disconnects and begins reconnection

Expected Behavior

• ClawTalk reconnects independently
• Slack message processing continues uninterrupted
• Channels are isolated from each other's connection failures

Actual Behavior

• ClawTalk reconnection logic consumes event loop cycles
• Slack Socket Mode client cannot ack events within the 3-second timeout
• Health monitor detects stale Slack socket and force-restarts the Slack provider
• Slack messages received during the gap are lost or delayed

Logs

16:37:47 [compaction-safeguard] new content uses 206.4% of context; dropped 3 older chunk(s)
16:43:28 [gateway] Pong timeout, closing connection
16:43:28 [gateway] ClawTalk disconnected: code=1006 reason=unknown
16:43:28 [gateway] ClawTalk reconnecting in 5s (attempt 1)
16:43:36 [gateway] ClawTalk connected, authenticating...
16:43:36 [gateway] ClawTalk authenticated (v0.2.0)
-- No Slack activity logged between 16:37 and manual restart at 17:13 --

After manual gateway restart, Slack reconnects and works normally.

Environment

• OpenClaw Gateway v2026.4.5
• Node.js v25.5.0
• Slack extension: Socket Mode
• ClawTalk extension: enabled, autoConnect: true
• Both extensions share agentId: "main"

Suggested Fix

One or more of:

• Run channel extensions in separate worker threads or child processes so one channel's WebSocket failure cannot starve another's event loop
• Ensure ClawTalk reconnection logic is fully non-blocking (no synchronous operations during reconnect)
• Decouple Slack ack processing from the main event loop (dedicated microtask queue)
• Add a configurable option to isolate channels into separate runtimes

Related Issues

• #56399 — Discord WebSocket crash kills gateway process (same class: one channel failure affects entire gateway)
• #39572 — stale-socket reconnect loop (related Slack reconnection behavior)
• #36687 — cross-channel reply leakage (channel isolation concern)