How does Kubernetes handle service discovery and load balancing internally among pods for Turing deployments?
