Question

A JVM (G1GC) microservice's CPU jumps from 35% to 95% after a deploy at 13:00, p99 latency doubles, but request throughput is unchanged. GC dashboards show GC CPU% climbing from 5% to 45%, allocation rate up from 200MB/s to 1.8GB/s, and young-GC frequency 5x higher; old-gen occupancy and full-GC count are normal. Heap size and traffic are unchanged. The 13:00 deploy added a new response-enrichment step that builds large intermediate collections per request. Triage and mitigate.