Summary

The Telegram channel plugin stops receiving messages after ~8 minutes when the long-polling connection times out. The plugin does not auto-reconnect, requiring manual gateway restarts to restore functionality.

Steps to reproduce

1. Configure OpenClaw with Telegram channel enabled (long-polling mode, not webhooks)
2. Start the gateway and send a test message (verify bot responds)
3. Wait ~8-10 minutes without interaction
4. Send another message to the bot
5. Observe that the bot does not respond
6. Check openclaw status - shows Telegram as "OK"
7. Check Telegram API getWebhookInfo endpoint - pending_update_count will be > 0

Expected behavior

The Telegram plugin should automatically reconnect when the long-polling connection times out, maintaining continuous message delivery without manual intervention.

Actual behavior

• Long-polling connection times out after ~500 seconds (~8.3 minutes)
• Plugin does not reconnect
• Messages queue on Telegram's side (visible as pending updates)
• openclaw status shows channel as "OK" (API is reachable, but polling loop is dead)
• Bot remains unresponsive until gateway is manually restarted

Environment

• OpenClaw version: 2026.1.30
• OS: macOS (likely affects all platforms)
• Install method: pnpm (global install)

Logs or screenshots

No error logs are generated when polling stops. The connection silently dies without logging a failure.

After adding retry policy to config:

{
  "channels": {
    "telegram": {
      "retry": {
        "attempts": 5,
        "minDelayMs": 1000,
        "maxDelayMs": 10000,
        "jitter": 0.3
      }
    }
  }
}

Result: 23+ hours continuous operation, zero timeouts, zero restarts required.

Proposed Fix

Make retry policy the default for Telegram channel configuration. This converts a hard failure (polling dies, manual restart needed) into a self-healing transient failure (exponential backoff reconnection).

Suggested Default Config

{
  "channels": {
    "telegram": {
      "retry": {
        "attempts": 5,
        "minDelayMs": 1000,
        "maxDelayMs": 10000,
        "jitter": 0.3
      }
    }
  }
}

This would make the Telegram integration production-ready out of the box without requiring users to implement external health monitoring scripts.