observer-effect-debugging

Race conditions and flaky tests often change behavior when observation alters execution timing. This Skill helps diagnose these probe-sensitive failures by classifying perturbation sources and distinguishing real system bugs from instrumentation side effects.

Debugging production-only race conditions requires comparing continuous, sampled, out-of-band, and post-hoc observation modes. This Skill guides you to capture failure evidence with minimal steady-state impact by reducing probe size and validating evidence.

To profile performance without affecting latency, you must reduce probe size and validate evidence with out-of-band corroboration. This Skill helps classify timing, cache, GC, and JIT effects to find trustworthy measurements during probe-sensitive latency investigations.

Fixing flaky tests that change under continuous monitoring involves comparing sampled versus continuous observation modes. This Skill helps classify timing and scheduling effects, enabling you to reduce probe impact and validate evidence through post-hoc corroboration.

Sampled monitoring is preferable over continuous profiling when continuous instrumentation alters system behavior. This Skill helps evaluate sampled-versus-continuous monitoring decisions by comparing observation modes to ensure you collect trustworthy evidence without inducing Heisenbugs.