When the pager goes off, the first ten minutes should not be a scavenger hunt

Where the time goes

The SRE team is measured on time to restore service, but most of that time is diagnosis: which of the afternoon's deploys, which commit, which pull request. The repair is quick once the change is known; finding the change is the slow part.

What changes when every alert is pre-correlated

Every alert arrives matched to the deployments in its blast radius, ranked, and resolved to the commit and PR. Diagnosis stops being a walk across four tools.

• Ranked candidate deploys, not a single confident guess that sends the team down the wrong path.
• If nothing shipped in the window, that is a finding in seconds: go look at infrastructure or traffic.
• Confirmed correlations raise change failure rate for the service and weigh on its next Gaja verdict.

What NovaPay refuses to claim in exchange

Time to a candidate root cause is not MTTR. ArcRelease shortens diagnosis, not the repair, and reports the two as separate numbers on separate charts. Conflating them would flatter the product and mislead the team.