Discord channel: stale-socket health-monitor restarts every 6 min on quiet guilds

openclaw version: 2026.4.20 (commit 115f05d) Pre-existing in: 2026.4.15 and earlier (pre-#43392 fix) #43392 fix: "delay HTTP bind until websocket handlers attached" — does NOT resolve this symptom

Summary

In a low-traffic Discord guild, the channel health monitor restarts the Discord client every ~6 min on a clockwork cadence, even though the bot is connected and functional. Cause: channelHealthCheckMinutes: 1 + channelStaleEventThresholdMinutes: 5

• no Discord events arriving from the guild within 5 min ⇒ stale-socket false positive.

Reproduction

• Single small Discord guild (~1 channel, low message volume)
• Default config:

  "gateway": {
    "channelHealthCheckMinutes": 1,
    "channelStaleEventThresholdMinutes": 5
  }

• Bot has correct intents enabled (Message Content Intent verified ON in dev portal)
• After bot connects, watch /tmp/openclaw/openclaw-YYYY-MM-DD.log

Observed cadence

2026-04-22T03:21:27 ready (6 plugins: ...; 15.7s)
2026-04-22T03:21:30 discord client initialized as ... (Open-claw); awaiting gateway readiness
2026-04-22T03:27:27 [discord:default] health-monitor: restarting (reason: stale-socket)
2026-04-22T03:33:27 [discord:default] health-monitor: restarting (reason: stale-socket)
...

Exactly 6 min between each restart (= 1 min health-check + 5 min stale threshold). Bot remains functional between restarts (messages send/receive correctly), but each restart cycles the entire plugin subsystem, triggering re-init of all extension plugins (Sendblue, semantic-memory, openclaw-web-search). On our system this also triggers harmless EADDRINUSE errors as Sendblue tries to re-bind port 3141 each cycle.

Root cause hypothesis

The health monitor counts only "Discord events" (messages, reactions, etc.) toward "socket alive" tracking. Discord's WS heartbeats and READY/RESUMED frames don't count toward the staleness clock, so any guild with <1 event per 5 min triggers false-positive restarts.

Suggested fix

Either:

1. Count WS heartbeat acks toward the freshness clock (most correct), OR
2. Increase the default channelStaleEventThresholdMinutes from 5 to 30+ minutes, OR
3. Document that quiet guilds should raise channelStaleEventThresholdMinutes (we set ours to 60 as a workaround, eliminating the loop)

Workaround applied locally

jq '.gateway.channelStaleEventThresholdMinutes = 60' \
  ~/.openclaw/openclaw.json > /tmp/new && mv /tmp/new ~/.openclaw/openclaw.json
supervisorctl restart openclaw-gateway

Result: no stale-socket restarts in the first hour after applying.

Environment

• LXC container, Debian 12, Node 22.22.2
• Single guild (K9's server), single channel (open-claw)
• Bot ID: 1478931832987844732, name Open-claw
• Message Content Intent: ON (verified in dev portal)
• Other channels (Telegram, Sendblue) on same gateway: no equivalent restart loop