Bug Report

Version: 2026.1.24-2 (d1dd8a1)
OS: Windows 10.0.26200 (x64), Node v24.12.0
Channel: Telegram

Problem

The gateway process crashes repeatedly due to unhandled promise rejections from failed etch calls. A single network hiccup kills the entire gateway.

Error

\
[clawdbot] Unhandled promise rejection: TypeError: fetch failed
at node:internal/deps/undici/undici:15845:13
at processTicksAndRejections (node:internal/process/task_queues:103:5)
\\

Frequency

Happened 3 times in one afternoon (Jan 29, 2026):

• 4:51 PM ET — crashed, was down for 1+ hour until manual restart
• 6:31 PM ET — crashed, watchdog restarted in 6s
• 6:33 PM ET — crashed again, watchdog restarted in 8s

Context

Gateway was running normally with Telegram channel active, multiple cron sessions, and browser automation sub-agents. The crash appears to occur during routine network operations (likely Telegram long-polling or API calls).

Expected Behavior

A failed fetch should be caught and retried or logged — not crash the entire gateway process. All outgoing network calls should be wrapped in try/catch or have .catch() handlers on their promises.

Workaround

Running a PowerShell watchdog script that polls the gateway every 60s and runs \clawdbot gateway start\ if unreachable. This limits downtime to ~10 seconds but doesn't fix the root cause.

Suggestion

Consider adding a global \process.on('unhandledRejection')\ handler that logs the error and continues rather than crashing, or audit all fetch/HTTP calls for proper error handling.