A rolling deploy is stuck at 50% and both versions are serving errors
A rolling deployment has replaced half the instances. Error rate is elevated on both old and new instances, not just the new ones. The rollout has stalled because health checks on the new pods are failing intermittently. Someone suggests completing the rollout so at least there is only one version.
The fix that looks right
Completing the rollout to "get to one version". It works: the errors stop within minutes, the graph recovers, and the incident looks resolved. What actually happened is that the compatibility break was resolved by eliminating the old version — which means the cache is now full of entries the previous release cannot read, and the rollback that was available five minutes ago no longer is. If the new version turns out to have a second, unrelated defect an hour later, the team discovers that their way back was spent on the first problem.