Regional error rate at 100%

You run active-active in three regions (us-east, us-west, eu-west). At 16:20, ~30% of users report total failure while others are fine. Dashboards: us-east error rate is 100%, the other two regions are at baseline; us-east's regional API gateway is returning connection-refused; a cloud-provider status page just posted a networking incident in us-east-1; your global DNS is still routing ~1/3 of traffic to us-east via latency-based routing. No deploy. How do you triage and mitigate?