Question

A blue-green cutover flips 100% to green at 09:00 for a service that calls a third-party geocoding API. Green is healthy and warmed in every internal sense. But at 09:00 the third-party API starts returning HTTP 429 to green, and ~10% of requests fail for 20 minutes before recovering. Context: the third-party rate-limits per source IP, and blue and green run in different subnets / behind different NAT egress IPs. Blue's egress IP had built up a higher allowed-burst allowance with the vendor over time; green's brand-new egress IP starts cold with the default lower limit. Dashboards: green's 429s spike at 09:00 then taper as the vendor's adaptive limit ramps up for the new IP; internal metrics all green. Triage, explain, and prevent.