Bug Report: Bonjour/ciao mDNS plugin causes repeated gateway crashes on headless Linux VPS ("CIAO PROBING CANCELLED")

Version: 2026.4.24 (and previous weeks)

Environment:

• Ubuntu 24.04 VPS (Contabo)
• Headless
• gateway.mode: local, bind: loopback
• Both Avahi daemon (avahi-daemon) and OpenClaw's bonjour plugin active → mDNS stack conflict
• avahi-daemon explicitly logs: *** WARNING: Detected another IPv4 mDNS stack running on this host. This makes mDNS unreliable...

Symptoms:

• Bonjour plugin watchdog gets stuck in probing / announcing state for ~12-13 seconds.
• Triggers unhandled promise rejection: CIAO PROBING CANCELLED or CIAO ANNOUNCEMENT CANCELLED.
• Gateway process exits with status 1.
• systemd restarts it in a tight loop.
• Internal WS probe fails (ECONNRESET / timeout).
• Multiple stability bundles created (see attached pattern).

Reproduction:

1. Run on a Debian/Ubuntu system that has Avahi installed (common on VPS).
2. Leave plugins.bonjour.enabled default (or even set to false — watchdog still runs).
3. Observe gateway logs + journalctl -u openclaw-gateway.service.

Workaround that works:

"plugins": {
  "entries": {
    "bonjour": {
      "enabled": false,
      "logLevel": "silent"
    }
  }
}

(Disabling the plugin + suppressing logs stops the crashes.)

Requested fix:

• Make the bonjour/ciao integration robust against other mDNS stacks (proper error handling on probing/announcement cancellation).
• Or add auto-disable logic when Avahi is detected.
• Prevent unhandled rejection from crashing the entire gateway.

Attached:

• Multiple stability bundles from /root/.openclaw/logs/stability/ (latest ones from 2026-04-26)
• journalctl excerpts showing the exact rejection
• openclaw security audit --deep output (shows related exposure warnings but not this bug)

This has been a recurring source of instability on this VPS.