Description

After upgrading to 2026.4.5, the gateway enters a crash-loop, restarting every ~60 seconds. The Control UI disconnects and reconnects with each restart.

Error

The following uncaught exception appears in ~/.openclaw/logs/gateway.err.log on every cycle:

[openclaw] Uncaught exception: Error: Attempted to reconnect zombie connection after disconnecting first (this shouldn't be possible)
    at Object.reconnectCallback (file:///opt/homebrew/lib/node_modules/openclaw/dist/provider-DR2mO1YM.js:13811:26)
    at Timeout.sendHeartbeat [as _onTimeout] (file:///opt/homebrew/lib/node_modules/openclaw/dist/provider-DR2mO1YM.js:13462:12)
    at listOnTimeout (node:internal/timers:605:17)
    at processTimers (node:internal/timers:541:7)

Behaviour

• Gateway starts successfully
• After ~60 seconds, the uncaught exception fires in the Telegram provider's heartbeat/reconnect callback
• Since it's uncaught, the Node.js process exits
• launchd (macOS, KeepAlive: true, ThrottleInterval: 1) immediately restarts the gateway
• The cycle repeats indefinitely
• The Control UI resets every ~60 seconds as a result

Steps to Reproduce

1. Have Telegram channel configured (channels.telegram.streaming: partial)
2. Upgrade to 2026.4.5
3. Start the gateway
4. Wait ~60 seconds — gateway crashes and restarts
5. Observe gateway.err.log for the uncaught exception

Environment

• OpenClaw version: 2026.4.5
• OS: macOS (arm64, Darwin 25.3.0)
• Node: v25.8.0
• Install method: Homebrew (/opt/homebrew/lib/node_modules/openclaw)
• Gateway mode: local, loopback
• Telegram streaming: partial
• Affected since: 2026.4.5 (confirmed working on 2026.4.2)

Workaround

Rolling back to 2026.4.2 resolves the issue:

npm install -g [email protected]

Related

Similar zombie/reconnect crash issues for Discord: #56588, #57291, #54752 — this appears to be the same class of bug now affecting the Telegram provider.