- Fix network_programming_latency metric reporting for Endpoints/EndpointSlice deletions, where we don't have correct timestamp ([#95363](https://github.com/kubernetes/kubernetes/pull/95363), [@wojtek-t](https://github.com/wojtek-t)) [SIG Network and Scalability]

statically to `IfNotPresent` for all required images. If you want to always pull the latest images before cluster init (as `Always` would do), run `kubeadm config images pull` before each `kubeadm init`. If you don't want the kubelet to pull any images at `kubeadm init` time, for example if you don't have an internet connection, you can also run `kubeadm config images pull` before `kubeadm init` or side-load the images some other way (such as `docker load -i image.tar`). Having the images locally cached...

- Don't prematurely close reflectors in case of slow initialization in watch based manager to fix issues with inability to properly mount secrets/configmaps. ([#104604](https://github.com/kubernetes/kubernetes/pull/104604), [@wojtek-t](https://github.com/wojtek-t))

* Add support for Azure internal load balancer. ([#45690](https://github.com/kubernetes/kubernetes/pull/45690), [@jdumars](https://github.com/jdumars))
* Fix `kubectl delete --cascade=false` for resources that don't have legacy overrides to orphan by default. ([#45713](https://github.com/kubernetes/kubernetes/pull/45713), [@kargakis](https://github.com/kargakis))

* Fix concurrent map access panic ([#65327](https://github.com/kubernetes/kubernetes/pull/65327), [@dashpole](https://github.com/dashpole))
    * Don't watch .mount cgroups to reduce number of inotify watches
    * Fix NVML initialization race condition
    * Fix brtfs disk metrics when using a subdirectory of a subvolume

* kubeadm: Fixes a small bug where `--config` and `--skip-*` flags couldn't be passed at the same time in validation. ([#49498](https://github.com/kubernetes/kubernetes/pull/49498), [@luxas](https://github.com/luxas))

  - Fix an error where existing nodes were not considered as destinations while finding place for pods in scale-down simulations
  - Remove redundant log lines and reduce severity around parsing kubeEnv
  - Don't treat nodes created by virtual kubelet as nodes from non-autoscaled node groups
  - Remove redundant logging around calculating node utilization
  - Add configurable `--network` and `--rm` flags for docker in `Makefile`

  * `daemonset-lookup-cache-size`

  Don't use these flags. Using deprecated flags causes the server to print a warning. Using a removed flag causes the server to abort the startup.

* The following deprecated flags were removed from `kubelet`:

  * `api-servers` - add apiserver addresses to the kubeconfig file instead.

- The `kube-proxy` command-line documentation was updated to clarify that
  `--bind-address` does not actually have anything to do with binding to an
  address, and you probably don't actually want to be using it. ([#120274](https://github.com/kubernetes/kubernetes/pull/120274), [@danwinship](https://github.com/danwinship))

- Fix instance not found issues when an Azure Node is recreated in a short time ([#93316](https://github.com/kubernetes/kubernetes/pull/93316), [@feiskyer](https://github.com/feiskyer)) [SIG Cloud Provider]
- Fix: don't use docker config cache if it's empty ([#92330](https://github.com/kubernetes/kubernetes/pull/92330), [@andyzhangx](https://github.com/andyzhangx)) [SIG Cloud Provider]