- Adds alpha metrics `apiserver_peer_proxy_errors_total` and `apiserver_peer_discovery_sync_errors_total` to apiserver to track errors encountered in peer proxying and peer discovery ([#137065](https://github.com/kubernetes/kubernetes/pull/137065), [@richabanker](https://github.com/richabanker)) [SIG API Machinery]

    cache.close()
    createNewCache()

    // The journal will have no record that 'a' was removed. It will have an entry for 'a', but when
    // it tries to read the cache files, it will find they were deleted. Once it encounters an entry
    // with missing cache files, it should remove it from the cache entirely.
    assertThat(cache.size()).isEqualTo(4)
    assertThat(cache["a"]).isNull()
    assertThat(cache.size()).isEqualTo(2)
  }

- `kubectl auth reconcile` now re-attempts reconciliation if it encounters a conflict error ([#133323](https://github.com/kubernetes/kubernetes/pull/133323), [@liggitt](https://github.com/liggitt)) [SIG Auth and CLI]

Some, such as `http2client` and `http2server`,
will be maintained much longer, even indefinitely.

When possible, each GODEBUG setting has an associated
[runtime/metrics](/pkg/runtime/metrics/) counter
named `/godebug/non-default-behavior/<name>:events`
that counts the number of times a particular program's
behavior has changed based on a non-default value
for that setting.
For example, when `GODEBUG=http2client=0` is set,

## Changelog since v1.18.1

## Changes by Kind

### Bug or Regression

- Client-go: resolves an issue with informers falling back to full list requests when timeouts are encountered, rather than re-establishing a watch. ([#89975](https://github.com/kubernetes/kubernetes/pull/89975), [@liggitt](https://github.com/liggitt)) [SIG API Machinery and Testing]

        *  Don't break metric storage when duplicate pod metrics encountered causing hpa to fail
* kube-apiserver: Fixed a regression accepting patch requests > 1MB ([#84963](https://github.com/kubernetes/kubernetes/pull/84963), [@liggitt](https://github.com/liggitt))

        if (!toolchainsResult.getProblems().isEmpty() && slf4jLogger.isWarnEnabled()) {
            slf4jLogger.warn("");
            slf4jLogger.warn("Some problems were encountered while building the effective toolchains");

            for (Problem problem : toolchainsResult.getProblems()) {
                slf4jLogger.warn("{} @ {}", problem.getMessage(), problem.getLocation());
            }

      apply {
        this.eventListenerFactory = eventListenerFactory
      }

    /**
     * Configure this client to retry or not when a connectivity problem is encountered. By default,
     * this client silently recovers from the following problems:
     *
     * * **Unreachable IP addresses.** If the URL's host has multiple IP addresses,

* API Registration is now in beta. ([#45247](https://github.com/kubernetes/kubernetes/pull/45247), [@mbohlool](https://github.com/mbohlool))

- Sometimes, the scheduler incorrectly placed a pod in the "unschedulable" queue instead of the "backoff" queue. This happened when some plugin previously declared the pod as "unschedulable" and then in a later attempt encounters some other error. Scheduling of that pod then got delayed by up to five minutes, after which periodic flushing moved the pod back into the "active" queue. ([#120334](https://github.com/kubernetes/kubernetes/pull/120334), [@pohly](https://github.com/pohly))...