Why does Kubernetes ignore failures… and then suddenly restart everything?

Not everything that looks like failure in Kubernetes is actually a failure. Sometimes it's just configuration (rules), not a real error. What confused me? While debugging a Pod, I saw this: Liveness