It was my mistake. I have two ingress-paths in my cluster; one for internal services, one for external. My own relay runs on the same cluster as hanami and was only serving traffic to external requests while cluster-internal requests couldn't even resolve the hostname of my relay. Now it's all sorte...