[BUG] Dispatch burns through API rate limit when OpenClaw gateway is unreachable

Preflight Checklist

• [x] I have searched existing issues and this hasn't been reported yet
• [x] This is a single bug report (please file separate reports for different bugs)
• [x] I am using the latest version of Claude Code

What's Wrong?

When the OpenClaw gateway (ai.openclaw.gateway) goes down, Dispatch continues retrying indefinitely instead of detecting the gateway is unreachable and stopping. This causes rapid API call consumption that exhausts the user's Anthropic API rate limit, leaving them unable to use Claude until the limit resets.

What Should Happen?

Dispatch should detect that the gateway is unreachable (e.g. via a health check before starting a task, or after N failed attempts) and fail gracefully with a clear error message rather than retrying indefinitely.

Error Messages/Logs

Steps to Reproduce

Stop or allow the OpenClaw gateway service to go down
Give Dispatch a task to execute
Dispatch begins attempting to process the task but cannot reach the gateway
Dispatch continues looping/retrying without stopping or surfacing an error
Anthropic API rate limit is hit

Claude Model

Sonnet (default)

Is this a regression?

I don't know

Last Working Version

_No response_

Claude Code Version

2.1.81 — you're up to date.

Platform

Anthropic API

Operating System

macOS

Terminal/Shell

Terminal.app (macOS)

Additional Information

macOS (Darwin 25.3.0)
OpenClaw gateway service: ai.openclaw.gateway (v2026.3.2)
Gateway port: 18789

Suggested fix

Add a gateway health check before Dispatch begins executing a task
Implement a circuit breaker or max retry cap so Dispatch stops after N consecutive failures
Surface a clear error to the user: "OpenClaw gateway is unreachable — please restart the service"
Environment