Tag
fault-localization
- Mar 20, 2026 Reproduction Fidelity and the Sources of Nondeterminism A failure that reproduces locally is not necessarily the failure that was observed, and the difference is a set of interleavings, clock readings, hash seeds, and externally timed events. observability
- Dec 10, 2025 Discriminating Tests and the Shape of a Suspect Set How a probe partitions the set of candidate causes, why an uninformative probe is worse than no probe, and how to price a test in bits against the time it takes. observability
- Oct 15, 2024 Reading a Diff for the Mechanism It Changed A bisected revision is a pointer to a change, not a cause. Separating default-value, ordering, and resource-lifecycle differences from cosmetic ones is what turns a pointer into a mechanism. observability
- Jul 16, 2024 Bisecting a Deploy Range When the Fault Is Not Monotone Binary search over versions and inputs assumes the symptom persists once introduced; what that assumption costs, how delta debugging generalises it, and where the search stops being worth running. observability