### Bug or Regression

- An inefficient lock in EndpointSlice controller metrics cache has been reworked. Network programming latency may be significantly reduced in certain scenarios, especially in clusters with a large number of Services. ([#107169](https://github.com/kubernetes/kubernetes/pull/107169), [@robscott](https://github.com/robscott)) [SIG Apps and Network]

  never enter the Succeeded phase may need to adapt their controllers. This is because as a consequence of 
  the change pods which use RestartPolicy=Always may end up in the Succeeded phase in two scenarios: pod 
  deletion and graceful node shutdown. ([#115331](https://github.com/kubernetes/kubernetes/pull/115331), [@mimowo](https://github.com/mimowo)) [SIG Cloud Provider, Node and Testing]

- The scheduling queue didn't notice any extenders' failures, it could miss some cluster events,
  and it could end up Pods rejected by Extenders stuck in unschedulable pod pool in 5min in the worst-case scenario.
  Now, the scheduling queue notices extenders' failures and requeue Pods rejected by Extenders appropriately. ([#122045](https://github.com/kubernetes/kubernetes/pull/122045), [@sanposhiho](https://github.com/sanposhiho)) [SIG Scheduling]

more information, see [#53356](https://issue.k8s.io/53356#issuecomment-333748618).

* A performance issue was identified in large-scale clusters when deleting thousands of pods simultaneously across hundreds of nodes. Kubelets in this scenario can encounter temporarily increased latency of `delete pod` API calls -- above the target service level objective of 1 second. If you run clusters with this usage pattern and if pod deletion latency could be an issue for you, you might want to wait...

- QueueingHint implementation for NodeAffinity is reverted because we found potential scenarios where events that make Pods schedulable could be missed. ([#122285](https://github.com/kubernetes/kubernetes/pull/122285), [@sanposhiho](https://github.com/sanposhiho)) [SIG Scheduling]

- An inefficient lock in EndpointSlice controller metrics cache has been reworked. Network programming latency may be significantly reduced in certain scenarios, especially in clusters with a large number of Services. ([#107091](https://github.com/kubernetes/kubernetes/pull/107091), [@robscott](https://github.com/robscott))

### Bug or Regression

- An inefficient lock in EndpointSlice controller metrics cache has been reworked. Network programming latency may be significantly reduced in certain scenarios, especially in clusters with a large number of Services. ([#107167](https://github.com/kubernetes/kubernetes/pull/107167), [@robscott](https://github.com/robscott)) [SIG Apps and Network]

### Bug or Regression

- An inefficient lock in EndpointSlice controller metrics cache has been reworked. Network programming latency may be significantly reduced in certain scenarios, especially in clusters with a large number of Services. ([#107168](https://github.com/kubernetes/kubernetes/pull/107168), [@robscott](https://github.com/robscott)) [SIG Apps and Network]

- QueueingHint implementation for NodeAffinity is reverted because we found potential scenarios where events that make Pods schedulable could be missed. ([#122327](https://github.com/kubernetes/kubernetes/pull/122327), [@sanposhiho](https://github.com/sanposhiho)) [SIG Scheduling]

- The scheduling queue didn't notice any extenders' failures, it could miss some cluster events,
  and it could end up Pods rejected by Extenders stuck in unschedulable pod pool in 5min in the worst-case scenario.
  Now, the scheduling queue notices extenders' failures and requeue Pods rejected by Extenders appropriately. ([#122043](https://github.com/kubernetes/kubernetes/pull/122043), [@sanposhiho](https://github.com/sanposhiho)) [SIG Scheduling]