# Self-Healing CI: Recovering a Proxy / Egress Connection Reset

> A connection reset through an egress proxy is a transient network blip, not a config bug. See the manual fix and how self-healing CI retries the request automatically.

Source: https://latchkey.dev/learn/self-healing-ci/self-healing-proxy-connection-reset  
Updated: 2026-06-26

A request reset on its way through an egress proxy hit a momentary network problem, not a broken proxy config -- the same request usually succeeds on a retry.

## What makes a failure safely retryable

Automatic retry is only correct for failures that are genuinely transient. Retrying a deterministic failure wastes minutes and hides a real defect, so the classification matters more than the retry mechanism.

- Safe to retry: network timeouts, registry 5xx, transient DNS failures, a service container that was not ready, a spot instance reclaimed mid-run.
- Not safe to retry: assertion failures, compile errors, lint violations, anything that fails identically on every attempt.
- Ambiguous, and worth investigating rather than retrying: out-of-memory kills, disk exhaustion, and flaky tests. These repeat under load and a retry only hides the trend.
- Always record that a retry happened. A pipeline that silently retries is a pipeline whose real failure rate you do not know.

## FAQ

### What causes Self-Healing CI: recovering a proxy / egress connection reset?

An outbound request fails with a connection reset (ECONNRESET) as it passes through an egress proxy or NAT gateway. The proxy config and the destination are fine; the connection was reset by a transient condition on the path. A human re-runs the job and the request completes unchanged.

### How do I fix Self-Healing CI: recovering a proxy / egress connection reset manually?

[object Object]

### Can Self-Healing CI: recovering a proxy / egress connection reset be fixed automatically?

A connection reset on the egress path has a recognizable transient signature, and the safe response is to retry with backoff. A self-healing CI pipeline detects the reset, retries the request, and only escalates if the destination stays unreachable, distinguishing a momentary egress blip from a real proxy or destination problem.

---

Latchkey runs CI/CD that repairs its own failures. Agent entry points: https://latchkey.dev/agent.txt, https://latchkey.dev/openapi.json, https://latchkey.dev/llms.txt