filtering on fields was incorrect. If watching objects with a filter, when an update was made that no longer matched the filter a DELETE event was correctly sent. However, the object that was returned by that delete was not the (correct) version before the update, but instead, the newer version. That meant the new object was not matched by the filter. This was a regression from behavior between cached watches on the server side and uncached watches, and thus broke downstream API clients. ([#46223](http...

  - The disk-cached discovery client is moved from k8s.io/client-go/discovery to k8s.io/client-go/discovery/cached/disk.
The memory-cached discovery client is moved from k8s.io/client-go/discovery/cached to k8s.io/client-go/discovery/cached/memory.
([#72214](https://github.com/kubernetes/kubernetes/pull/72214), [@caesarxuchao](https://github.com/caesarxuchao))

  - Regenerate list of EC2 instances
  - Fix pricing endpoint in AWS China Region
  
  ### Azure
  - Add optional jitter on initial VMSS VM cache refresh, keep the refreshes spread over time
  - Serve from cache for the whole period of ongoing throttling
  - Fix unwanted VMSS VMs cache invalidations
  - Enforce setting the number of retries if cloud provider backoff is enabled
  - Don't update capacity if VMSS provisioning state is updating

filtering on fields was incorrect.  If watching objects with a filter, when an update was made that no longer matched the filter a DELETE event was correctly sent.  However, the object that was returned by that delete was not the (correct) version before the update, but instead, the newer version.  That meant the new object was not matched by the filter.  This was a regression from behavior between cached watches on the server side and uncached watches, and thus broke downstream API clients. ([#46223](ht...

- kubeadm: added "disable success" and "disable denial" as parameters of the "cache" plugin in the Corefile managed by kubeadm. This is to prevent conflicting responses during CoreDNS cache updates. ([#128359](https://github.com/kubernetes/kubernetes/pull/128359), [@matteriben](https://github.com/matteriben)) [SIG Cluster Lifecycle]

- Clear pod binding cache on bind error to make sure stale pod binding cache will not be used. ([#71212](https://github.com/kubernetes/kubernetes/pull/71212), [@cofyc](https://github.com/cofyc))
- Fixed a scheduler panic due to internal cache inconsistency ([#71063](https://github.com/kubernetes/kubernetes/pull/71063), [@Huang-Wei](https://github.com/Huang-Wei))

- Kube-scheduler now retries scheduling Pods rejected by the PodTopologySpread plugin when related Pods are created, deleted, updated, or when a node matches the specified topologyKey. ([#122195](https://github.com/...

- Fixed an issue where the kubelet token cache returned stale tokens when service accounts were recreated with the same name. The cache is now UID-aware. Additionally, the new `TokenRequestServiceAccountUIDValidation` feature gate (Beta, enabled by default) ensures the `TokenRequest` UID matches the service account UID when set. ([#132803](https://github.com/kubernetes/kubernetes/pull/132803), [@aramase](https://github.com/aramase))...

- Supporting Cross RG resources (disks, Azure File and node [Experimental]

Besides the above new features, support for Azure Virtual Machine Scale Sets (VMSS) and Cluster-Autoscaler is now stable and considered GA:

- Azure virtual machine scale sets (VMSS) allow you to create and manage identical load balanced VMs that automatically increase or decrease based on demand or a set schedule.

- To reduce chances of throttling, VM cache is set to nil when Azure node provisioning state is deleting ([#87635](https://github.com/kubernetes/kubernetes/pull/87635), [@feiskyer](https://github.com/feiskyer)) [SIG Cloud Provider]