* CRI: `kubectl logs -f` now stops following when container stops, as it did pre-CRI. ([#44406](https://github.com/kubernetes/kubernetes/pull/44406), [@Random-Liu](https://github.com/Random-Liu))
* Azure cloudprovider: Reduce the polling delay for all azure clients to 5 seconds. This should speed up some operations at the cost of some additional quota. ([#43699](https://github.com/kubernetes/kubernetes/pull/43699), [@colemickens](https://github.com/colemickens))

- `kube-apiserver` updated:

* [fluentd-gcp addon] Increase CPU limit for fluentd to 1 core to achieve 100kb/s throughput. ([#62430](https://github.com/kubernetes/kubernetes/pull/62430), [@bmoyles0117](https://github.com/bmoyles0117))
* GCE: Fixed operation polling to adhere to the specified interval. Furthermore, operation errors are now returned instead of ignored. ([#64630](https://github.com/kubernetes/kubernetes/pull/64630), [@nicksardo](https://github.com/nicksardo))

- Potentially breaking change - Updating the polling interval for Windows stats collection from 1 second to 10 seconds ([#116546](https://github.com/kubernetes/kubernetes/pull/116546), [@marosset](https://github.com/marosset)) [SIG Node and Windows]

## Changelog since v1.35.0

## Changes by Kind

### Feature

- Kubeadm: when patching a Node object do not exit early on unknown (non-allowlisted) API errors. Instead, always retry within the duration of the polling for getting and patching a Node object. ([#136072](https://github.com/kubernetes/kubernetes/pull/136072), [@neolit123](https://github.com/neolit123)) [SIG Cluster Lifecycle]

    * If your config files contains fields with wrong case, the config files will be now invalid.
* GCE: Fixes operation polling to adhere to the specified interval. Furthermore, operation errors are now returned instead of ignored. ([#64630](https://github.com/kubernetes/kubernetes/pull/64630), [@nicksardo](https://github.com/nicksardo))

## Changelog since v1.33.7

## Changes by Kind

### Feature

- Kubeadm: when patching a Node object do not exit early on unknown (non-allowlisted) API errors. Instead, always retry within the duration of the polling for getting and patching a Node object. ([#136070](https://github.com/kubernetes/kubernetes/pull/136070), [@neolit123](https://github.com/neolit123)) [SIG Cluster Lifecycle]

## Changelog since v1.34.3

## Changes by Kind

### Feature

- Kubeadm: when patching a Node object do not exit early on unknown (non-allowlisted) API errors. Instead, always retry within the duration of the polling for getting and patching a Node object. ([#136071](https://github.com/kubernetes/kubernetes/pull/136071), [@neolit123](https://github.com/neolit123)) [SIG Cluster Lifecycle]

## Changelog since v1.32.11

## Changes by Kind

### Feature

- Kubeadm: when patching a Node object do not exit early on unknown (non-allowlisted) API errors. Instead, always retry within the duration of the polling for getting and patching a Node object. ([#136069](https://github.com/kubernetes/kubernetes/pull/136069), [@neolit123](https://github.com/neolit123)) [SIG Cluster Lifecycle]

An alternative implementation of the `CronJob` controller is now available as an alpha feature in this release, which has experimental performance improvement by using informers instead of polling. While this will be the default behavior in the future, you can [try them in this release through a feature gate](https://docs.k8s.io/concepts/workloads/controllers/cron-jobs/).

### PID Limits graduates to General Availability