Error rate spike after deploy

You're on call for a single web API. Ten minutes ago the error-rate dashboard jumped from ~0.1% to ~40% of requests returning HTTP 500, and latency on the success path is unchanged. The deploy timeline shows a new version went out 12 minutes ago; nothing else changed (no traffic spike, no dependency alerts). The on-call channel is filling up with customer reports. How do you triage and mitigate?