Whenever you're ready, walk me through how you'd approach this.
How do Kubernetes components like etcd, the API server, and controllers work under the hood?
Explain the complete lifecycle of a workload change, from a client request through authentication, admission, persistence, watches, reconciliation, scheduling, and kubelet execution. Demonstrate how you would investigate stale state, failed watches, controller retries, and an unavailable etcd cluster. Include relevant commands or code, data consistency guarantees, and monitoring signals.