instance, which is a good example. It refactors `IndicesRequestCache` in such
a way that:
 - it no longer depends on objects that are hard to instantiate such as
   `IndexShard` or `SearchContext`,
 - time-based eviction is applied on top of the cache rather than internally,
   which makes it easier to assert on what the cache is expected to contain at
   a given time.

=== Bad practices

==== Use randomized-testing for coverage

- The `pod/eviction` subresource now accepts `policy/v1` eviction requests in addition to `policy/v1beta1` eviction requests ([#100724](https://github.com/kubernetes/kubernetes/pull/100724), [@liggitt](https://github.com/liggitt))

- Fixed two regressions introduced by the `PodDisruptionConditions` feature (on by default in 1.26):
  - pod eviction API calls returned spurious precondition errors and required a second evict API call to succeed

* Fix metrics API group name in audit configuration ([#53493](https://github.com/kubernetes/kubernetes/pull/53493), [@piosz](https://github.com/piosz))
* Use separate client for leader election in scheduler to avoid starving leader election by regular scheduler operations. ([#53793](https://github.com/kubernetes/kubernetes/pull/53793), [@wojtek-t](https://github.com/wojtek-t))

    filesystem.setFaultyDelete(cacheDir / "a.0", true)
    taskFaker.runNextTask()

    // Confirm we prevent edits after a trim failure.
    assertThat(cache.edit("c")).isNull()

    // A successful eviction should allow new writes.
    filesystem.setFaultyDelete(cacheDir / "a.0", false)
    cache.evictAll()
    set("c", "cc", "cc")
    assertValue("c", "cc", "cc")
  }

  @Test

* Kubectl
  * Display line number on JSON errors
  * Add flag -t as shorthand for --tty
* Resources
  * Improved node stability by *optionally* evicting pods upon memory pressure - [Design Doc](https://github.com/kubernetes/kubernetes/blob/release-1.3/docs/proposals/kubelet-eviction.md)
  * **alpha**: NVIDIA GPU support ([#24836](https://github.com/kubernetes/kubernetes/pull/24836), [@therc](https://github.com/therc))

- DRA device taint eviction: the controller might have reported "1 pod needs to be evicted in 1 namespace. 1 pod evicted since starting the controller." when only a single pod is involved, depending on timing (pod evicted, informer cache not updated yet). It would eventually...

- Fixed the bug where `spec.terminationGracePeriodSeconds` of the pod will always be overwritten by the MaxPodGracePeriodSeconds of the soft eviction, you can enable the `AllowOverwriteTerminationGracePeriodSeconds` feature gate, which will restore the previous behavior.  If you do need to set this, please file an issue with the Kubernetes project to help contributors understand why you needed...

* Use separate client for leader election in scheduler to avoid starving leader election by regular scheduler operations. ([#53793](https://github.com/kubernetes/kubernetes/pull/53793), [@wojtek-t](https://github.com/wojtek-t))

- Kubectl/drain: add disable-eviction option.Force drain to use delete, even if eviction is supported. This will bypass checking PodDisruptionBudgets, and should be used with caution. ([#85571](https://github.com/kubernetes/kubernetes/pull/85571), [@michaelgugino](https://github.com/michaelgugino)) [SIG CLI]...