OpenClaw 2026.5.12 isolated Telegram ingress can wedge inbound delivery while the poller still appears healthy.

Observed on local install:

• isolated Telegram worker wrote telegram/ingress-spool-default/0000000379952079.json
• update-offset-default.json advanced to lastUpdateId=379952079
• main gateway did not dispatch the spooled update into the session
• Telegram would not redeliver because offset was already advanced
• spool drain path awaited bot.handleUpdate(update.update) without a timeout/liveness watchdog, so drainActive could stay true indefinitely and future drain ticks no-op

Local mitigation applied to installed runtime:

• added deadline around spooled bot.handleUpdate
• added stuck-drain diagnostics
• restarted gateway; stuck spool drained and inbound Telegram resumed

The durable fix should address isolated ingress offset/dispatch semantics and add watchdog/stale-spool diagnostics in source.