Inside loops, time.After creates a new timer goroutine each iteration
that cannot be GC'd until it fires. In long-running loops (supervisor
restart backoff, Telegram polling, restart-context polling, CP stop
retry) this leaks goroutines proportional to iteration count.
Replace with time.NewTimer + timer.Stop() on ctx cancellation so the
timer is cleaned up immediately when the goroutine exits.
Affected files:
- supervised/supervised.go (RunWithRecover backoff)
- channels/telegram.go (429 retry + poll error sleep)
- handlers/restart_context.go (online + heartbeat polling)
- handlers/workspace_restart.go (cpStop retry backoff)
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Molecule AI Dev Engineer A (Kimi)
02a3de7c0e