Summary

Bug Report

Summary

When a config.patch (or any action that triggers a SIGUSR1 gateway restart) is applied during an active Telegram session, any in-flight agent responses are silently lost with no retry mechanism.

Steps to reproduce

1. Have an active Telegram conversation with the bot
2. Apply a config.patch while an agent turn is in progress (message being generated)
3. Agent finishes generating the reply, but Telegram sendMessage fails

Expected behavior

• Gateway should buffer/queue outgoing messages during restart
• Or retry sendMessage after reconnecting
• Or send a "I restarted, please resend" notification on reconnect

Actual behavior

Message silently lost — user never receives it

• No retry attempted, no fallback

telegram sendMessage failed: Network request for 'sendMessage' failed!
telegram final reply failed: HttpError: Network request for 'sendMessage' failed!

OpenClaw version

2026.2.17

Operating system

Ubuntu 22.04.5 LTS (kernel 5.15.0-170)

Install method

No response

Logs, screenshots, and evidence

Extracted from /tmp/openclaw/openclaw-2026-02-20.log at 2026-02-21T02:53:47 UTC:

{"subsystem":"gateway/channels/telegram","message":"telegram sendMessage failed: Network request for 'sendMessage' failed!","time":"2026-02-21T02:53:47.527Z"}

{"subsystem":"gateway/channels/telegram","message":"telegram final reply failed: HttpError: Network request for 'sendMessage' failed!","time":"2026-02-21T02:53:47.532Z"}

Timeline:
- 02:52:43 UTC — Agent turn started (messageChannel=telegram)
- 02:53:36 UTC — config.patch applied → SIGUSR1 restart triggered  
- 02:53:47 UTC — Agent reply ready, sendMessage fails (Telegram connection dropped during restart)
- Result: user never received the response

Impact and severity

No response

Additional information

Environment

• OpenClaw: 2026.2.17
• Channel: Telegram
• Trigger: config.patch → SIGUSR1 restart (~3-5s downtime)

Impact

High — users lose responses with no indication anything went wrong.