Bug Report for Clawdbot Team

Issue: Unhandled Promise Rejection causing Gateway crash

Environment

• Version: 2026.1.24-3
• OS: Windows_NT 10.0.26200 (x64)
• Node: v24.0.1
• Model: anthropic/claude-sonnet-4-5
• Gateway Mode: local (loopback)
• Channel: Telegram (enabled)

Error Details

[clawdbot] Unhandled promise rejection: TypeError: fetch failed
    at node:internal/deps/undici/undici:15363:13
    at processTicksAndRejections (node:internal/process/task_queues:105:5)

Timestamp: 2026-01-29T19:08:14.678Z
Log File: \tmp\clawdbot\clawdbot-2026-01-29.log (line ~16MB in)

Context

• State: Gateway was idle (no active agent runs)
• Last Activity: Normal WebSocket polling (node.list) at 19:08:11
• Crash: 3 seconds later at 19:08:14
• Recovery: Auto-restart via Windows Task Scheduler successful

Activity before crash:

• 19:08:06 - node.list response (3ms)
• 19:08:11 - node.list response (4ms)
• 19:08:14 - 💥 Crash

No incoming Telegram messages or user activity at crash time.

Suspected Cause

Unhandled rejection in background fetch operation, likely:

1. Telegram Bot API long-polling (getUpdates) timeout/network failure
2. Heartbeat check to external service
3. Stale HTTP request cleanup

The crash occurs during idle time with no active agent workload, suggesting a background polling or health-check operation.

Expected Behavior

Network failures in background operations should be caught and logged, not crash the gateway process.

Suggested Fix

Add .catch() handlers to all background fetch operations:

• Telegram getUpdates polling
• Heartbeat/health checks
• Any cleanup/timeout handlers

Consider adding a global unhandled rejection handler with graceful degradation instead of process crash.

Reproduction

Intermittent - appears to happen during network instability or API timeouts during idle periods. Not reliably reproducible.

Impact

Severity: Medium

• Gateway crashes and requires restart
• Auto-recovery works (Task Scheduler restart)
• No data loss
• User experience interrupted briefly