architect-diagnose-performance

Diagnose latency and throughput regressions by comparing baseline versus incident metrics across affected endpoints, correlating observability signals with recent commits to localize the bottleneck stage and generate a safe remediation plan.

Triage performance incidents by identifying impacted endpoints and incident windows, then comparing baseline versus incident metrics to infer the bottleneck stage and generate a remediation plan with a confidence score.

Correlating recent code changes with observability data identifies performance bottlenecks by matching latency, error distributions, and machine health trends against specific commits to determine the likely cause of saturation or throughput drops.

Yes, this approach supports post-regression analysis on services suffering capacity pressure by applying baseline-versus-incident metric comparisons to produce a bottleneck hypothesis with confidence and a targeted remediation plan.

The smallest safe intervention for fixing a performance regression is a remediation plan focused on the most localized bottleneck stage, capturing relevant metrics and confidence levels to restore capacity without broad changes.

Baseline versus incident metric comparisons are required to localize a bottleneck because they reveal the exact route or service stage where latency, saturation, or errors diverged from normal operating conditions to pinpoint the regression cause.