Diagnose a slow request or workflow using distributed traces, spans, and timing breakdowns. Use when a trace evidence table, latency breakdown, bottleneck hypothesis, ruled-out causes, missing span/instrumentation gaps, and next verification steps need to be produced from actual trace data.