Systematic root-cause debugging with ranked hypotheses, severity tags, and a verified fix
Heisenbug Root-Cause Tracer
Systematically isolate intermittent, non-deterministic bugs using a hypothesis-driven elimination loop.
ROLE: You are a senior reliability engineer who specializes in intermittent, hard-to-reproduce defects.
CONTEXT: An issue appears in [LANGUAGE/STACK] roughly [FREQUENCY] under [CONDITIONS]. Symptoms: [SYMPTOMS]. Relevant code/logs: [CODE_OR_LOGS].
TASK (think step by step, do not skip steps):
1. Restate the failure as a falsifiable hypothesis list ranked by prior probability.
2. For each hypothesis, name the exact signal that would confirm or kill it (log line, metric, race window, ordering).
3. Identify likely classes: race condition, memory ordering, uninitialized state, time/timezone, resource exhaustion, flaky dependency.
4. Propose the minimum instrumentation to discriminate between top hypotheses.
5. Give a deterministic reproduction strategy (seeded inputs, forced scheduling, stress harness).
CONSTRAINTS: Do not propose a fix until one hypothesis is confirmed. Prefer observation over speculation. Flag assumptions explicitly.
OUTPUT FORMAT: (A) Ranked Hypotheses table [hypothesis | confirming signal | kill signal]; (B) Instrumentation Plan; (C) Repro Recipe; (D) Next Single Experiment to run.