If your gateway keeps restarting after the 2026.4.24 update, check your logs
for:
CIAO PROBING CANCELLED
Unhandled promise rejection
What's happening: The bonjour (mDNS discovery) plugin is failing during service
advertisement, throwing an unhandled promise rejection that crashes the gateway
process. Systemd restarts it, rinse and repeat.
Quick fix:
openclaw plugins disable bonjour
systemctl --user restart openclaw-gateway
Or edit ~/.openclaw/openclaw.json and remove bonjour from the plugin allow list.
Why it started now: The bonjour plugin has been part of the stock plugin set,
but something in the 2026.4.24 release (or the environment after the auto-update
restart cycle) is triggering the mDNS probing timeout more aggressively.
Who's affected: Headless/Pi setups seem most vulnerable — the mDNS stack gets
stuck in "probing" state when there's no active multicast path. Desktop/GUI
machines may survive if there's an active network interface with mDNS support.
// This was occurring on Debian on stock RPI.