* kubectl: Fix bug that showed terminated/evicted pods even without `--show-all`. ([#48786](https://github.com/kubernetes/kubernetes/pull/48786), [@janetkuo](https://github.com/janetkuo))
* Never prevent deletion of resources as part of namespace lifecycle ([#48733](https://github.com/kubernetes/kubernetes/pull/48733), [@liggitt](https://github.com/liggitt))

- Terminating a restartPolicy=Never pod no longer has a chance to report the pod succeeded when it actually failed. ([#88440](https://github.com/kubernetes/kubernetes/pull/88440), [@smarterclayton](https://github.com/smarterclayton)) [SIG Node and Testing]

* Try self-repair scheduler cache or panic ([#37379](https://github.com/kubernetes/kubernetes/pull/37379), [@wojtek-t](https://github.com/wojtek-t))
* Ignore mirror pods with RestartPolicy == Never in restart tests ([#34462](https://github.com/kubernetes/kubernetes/pull/34462), [@yujuhong](https://github.com/yujuhong))

## Changelog since v1.5.1

### Other notable changes

* Fixes NotAuthenticated errors that appear in the kubelet and kube-controller-manager due to never logging in to vSphere ([#36169](https://github.com/kubernetes/kubernetes/pull/36169), [@robdaemon](https://github.com/robdaemon))

        if you see such issues, please use the legacy optimizer, and file
        an issue in the Keras GitHub repo.

    The old Keras optimizer will never be deleted, but will not see any
    new feature additions.
    New optimizers (e.g., `Adafactor`) will
    only be implemented based on `tf.keras.optimizers.Optimizer`, the new
    base class.

	// insufficient; if the user specifies CGO_CFLAGS=-fdiagnostics-color,
	// GCC will ignore TERM, and GCC can also be configured at compile-time
	// to ignore TERM.)
	stderr := p.gccErrors(b.Bytes(), "-fdiagnostics-color=never")
	if strings.Contains(stderr, "unrecognized command line option") {
		// We're using an old version of GCC that doesn't understand
		// -fdiagnostics-color. Those versions can't print color anyway,

- The pod phase lifecycle guarantees that terminal Pods, those whose states are Unready or Succeeded, can not regress and will have all container stopped. Hence, terminal Pods will never be reachable and should not publish their IP addresses on the Endpoints or EndpointSlices, independently of the Service TolerateUnready option. ([#110260](https://github.com/kubernetes/kubernetes/pull/110260), [@robscott](https://github.com/robscott))...

* Advanced audit allows logging failed login attempts ([#51119](https://github.com/kubernetes/kubernetes/pull/51119), [@soltysh](https://github.com/soltysh))
* kubeadm: Add support for using an external CA whose key is never stored in the cluster ([#50832](https://github.com/kubernetes/kubernetes/pull/50832), [@nckturner](https://github.com/nckturner))

- Fixed a bug where the Kubelet miscalculated the process usage of pods, causing pods to never get evicted for PID usage. ([#124101](https://github.com/kubernetes/kubernetes/pull/124101), [@haircommander](https://github.com/haircommander)) [SIG Node and Testing]

            (candidateVersion.isSnapshot && candidateVersion < gradleVersion)
                || candidateVersion.version.endsWith("-snapshot-1")
        }

        // Remove state for old versions of Gradle that we're unlikely to ever require again
        fileSystemOperations.removeOldVersionsFromDir(workerDir.dir("caches"), expireDistributionCache)

        // Remove scripts caches