Repair Circuit Breaker State Races

The problem

Circuit breakers often suffer from race conditions where stale callbacks, non-atomic transitions, or duplicate probes corrupt the internal state. This leads to breakers staying open after recovery, closing due to outdated successes, or admitting multiple probes despite a single-probe contract.

What it does

• Enforces a strict generation-based contract to reject stale callbacks and timer updates.
• Makes half-open probe admission atomic within the existing repository concurrency model.
• Fixes state leaks where cancellation or timeouts leave the breaker in an inconsistent generation.
• Replaces flaky wall-clock sleeps in tests with deterministic clocks and barriers.
• Implements generation-aware metrics to track transition reasons and rejected callbacks.

Frameworks & tools

Works with custom or wrapped state machines in environments using Read, Edit, and Bash tools. Applicable to languages with async/concurrent primitives like Node.js, Go, or Java.

Why this beats prompting it yourself

General prompts often miss the nuances of epoch-based state management, leading to "fixes" that just narrow the race window. This skill applies a formal ownership register to every transition, ensuring that only the current breaker generation can mutate state.

Use cases

• Preventing a stale success from closing a breaker that has already moved to a new failure epoch.
• Stopping multiple probes from firing simultaneously during a half-open state.
• Ensuring manual resets do not race with in-flight probe completions or old timers.
• Converting non-deterministic circuit breaker tests into reliable, clock-controlled assertions.

Known limitations

Does not modify thresholds, retry budgets, or fallback policies. Will not introduce distributed coordination if no existing contract exists in the repository.