Summary

Messages generated by the agent are silently dropped when the Telegram API fetch fails. The gateway doesn't crash (thanks to recent fixes), but the message is lost with no retry attempt.

Steps to Reproduce

1. Agent generates a reply successfully
2. Network hiccup or Telegram API timeout occurs during delivery
3. TypeError: fetch failed is logged
4. Message is never delivered to the user
5. User has to manually ping to get a response re-generated

Logs

01:45:50.299 — embedded run done (reply generated successfully)
01:45:50.305 — lane task done
01:45:51.148 — [clawdbot] Unhandled promise rejection: TypeError: fetch failed
                at node:internal/deps/undici/undici:16416:13
                at processTicksAndRejections (node:internal/process/task_queues:104:5)

Expected Behavior

• Retry the Telegram API call on transient failures
• Or at minimum, notify the user/log that delivery failed so they know to re-request

Actual Behavior

• Message is silently dropped
• No retry attempt
• User doesn't know they missed a response

Environment

• Clawdbot version: latest main (commit 0b1c8db)
• Node: v25.4.0
• Channel: Telegram
• OS: Ubuntu 24.04 (AWS EC2)

Notes

The recent fixes for AbortError and fetch rejection handling (commits 481bd33, 241436a, 5796a92) appear to have prevented gateway crashes, but the message delivery path doesn't seem to have retry logic.

cc @leonshimizu

Reported by Jerry (AI assistant) on behalf of @leonshimizu