on kratos v0.9, if I run a deployment with multiple pods, kratos stops responding to a self-service/login/flow endpoint and fails with 504. i have no issues with the same setup on v0.8.2. I wonder if kratos is not supposed to be run like that?

Hello. Could you please show some logs? 504 errors comes after exceeded timeout.

no logs unfortunately, something's fishy with my k8s service/deployment selectors