openclaw - 💡(How to fix) Fix [Bug] Isolated cron sessions still cannot deliver to Telegram on v2026.3.28 [1 participants]

• Telegram outbound adapter loads at gateway startup and works initially
• After ~12h the adapter silently dies (possibly due to IPv6 ETIMEDOUT/ENETUNREACH on NAT-only VM)
• All isolated cron deliveries fail with Outbound not configured for channel: telegram
• Jobs with bestEffort: true report ok + delivered: false; jobs without it go to error state
• Gateway process stays alive — systemd sees no problem
• openclaw doctor reports Telegram: ok even while deliveries are failing
• On gateway restart, delivery-recovery successfully delivers pending messages

Fix Plan

To resolve the issue of Telegram outbound adapter silently dying after ~12 hours, we need to ensure that isolated cron jobs have access to the gateway's outbound context.

Here are the steps to fix the issue:

• Change the sessionTarget of all cron jobs from "isolated" to a named session, e.g., "session:cron-<job-name>".
• This will allow the cron jobs to access the gateway's outbound context and prevent the Telegram adapter from dying.

Example code snippet:

// Before
{
  "sessionTarget": "isolated",
  "delivery": {
    "mode": "announce",
    "channel": "telegram",
    "to": "<userId>"
  }
}

// After
{
  "sessionTarget": "session:cron-my-job",
  "delivery": {
    "mode": "announce",
    "channel": "telegram",
    "to": "<userId>"
  }
}

Verification

To verify that the fix worked:

• Restart the gateway and check that the Telegram adapter is loaded and working correctly.
• Run the cron jobs and check that they are delivering messages successfully.
• Monitor the gateway logs for any errors related to the Telegram adapter or outbound deliveries.

Extra Tips

• Make sure to update all cron jobs to use named sessions to prevent similar issues in the future.
• Consider implementing additional logging and monitoring to detect and respond to similar issues promptly.