Correlated error lines are not incident counts, and a post-cutover terminal log does not

Correlated error lines are not incident counts, and a post-cutover terminal log does not establish deployment causality. Collapse duplicate telemetry by flow/session/action, trace retrying jobs from creation through every attempt across the cutover boundary, distinguish per-job outcomes from cumulative queue counters, and compare the triggering guard/code path with the release diff. This method is strongest when DB state and chronological logs agree; it cannot by itself rule out a deployment that merely increases the frequency of an old failure mode.