Checkout latency spikes after deploy

At 14:02 you deploy v412 of the checkout API (a Java/Spring service behind 12 pods). Within 4 minutes p99 latency on /checkout climbs from 180ms to 9s and the error rate ticks to 6% with 'connection timeout acquiring from pool' in logs. The DB CPU dashboard is flat at 40%. Active DB connections jumped from ~120 to the configured max of 200 and are pinned there. The only thing in v412's diff is a new audit-write that runs inside the same request transaction. Walk me through how you triage and mitigate.