Bug type

Stability bug (crash)

Summary

When the host network is temporarily unreachable (e.g., ISP hiccup, DNS timeout), the SSRF/Telegram module in the Gateway throws an unhandled EHOSTUNREACH error that crashes the entire Gateway process.

Steps to reproduce

1. Run OpenClaw Gateway as a LaunchAgent service
2. Experience a temporary network interruption (ISP-level, not just Telegram API outage)
3. Observe: Gateway crashes with uncaught exception in SSRF module

Expected behavior

Gateway should catch EHOSTUNREACH and other transient network errors gracefully — log the error, back off, and retry — without crashing the process.

Actual behavior

Gateway process crashes. Requires LaunchAgent KeepAlive: true to auto-restart, which creates a brief service interruption for all connected channels.

Environment

• OpenClaw: 2026.4.23 (a979721)
• OS: macOS 26.3 (arm64)
• Install: Homebrew package
• Channel: Telegram (likely affects all channels using SSRF)

Logs

Error: connect EHOSTUNREACH <ip>:443
    at TCPConnectWrap.afterConnect [as oncomplete] (net.js:... )

(Full stack trace involves the SSRF module's fetch/request path to Telegram API)

Impact

Any transient network hiccup causes a full Gateway restart, disrupting all active sessions and in-progress tool calls. With LaunchAgent watchdog this auto-recovers, but the interruption is non-trivial during active work.