* --proxy-mode=ipvs
    * --cluster-cidrs=<cidr>,<cidr>
* The apiserver now uses http/1.1 to communicate with admission webhooks, opening multiple connections to satisfy concurrent requests, and allowing spreading requests across multiple backing pods. ([#82090](https://github.com/kubernetes/kubernetes/pull/82090), [@liggitt](https://github.com/liggitt))

line-height: 1.5;\n  border-radius: 0.2rem;\n}\n\n.form-control-lg {\n  height: calc(1.5em + 1rem + 2px);\n  padding: 0.5rem 1rem;\n  font-size: 1.25rem;\n  line-height: 1.5;\n  border-radius: 0.3rem;\n}\n\nselect.form-control[size], select.form-control[multiple] {\n  height: auto;\n}\n\ntextarea.form-control {\n  height: auto;\n}\n\n.form-group {\n  margin-bottom: 1rem;\n}\n\n.form-text {\n  display: block;\n  margin-top: 0.25rem;\n}\n\n.form-row {\n  display: -ms-flexbox;\n  display: flex;\n  -ms-flex-wrap:...

### Feature

- 'Added selector validation to `HorizontalPodAutoscaler`: when multiple HPAs
  select the same set of Pods, scaling now will be disabled for those HPAs with
  the reason `AmbiguousSelector`. This change also covers a case when multiple HPAs
  point to the same deployment.'

- Fixed a regression since 1.24 in the scheduling framework when overriding MultiPoint plugins (e.g. default plugins).
  The incorrect loop logic might lead to a plugin being loaded multiple times, consequently preventing any Pod from being scheduled, which is unexpected. ([#122366](https://github.com/kubernetes/kubernetes/pull/122366), [@caohe](https://github.com/caohe)) [SIG Scheduling]

                th.close();
            } else {
                // Let connection pooling handle cleanup - this improves performance
                // for scenarios where the same file or directory is accessed multiple times
                log.trace("Tree handle released to connection pool for reuse");
                // The connection pool will handle cleanup based on its own timeout settings
            }
        }

- Fix Windows kube-proxy (winkernel) issue where stale RemoteEndpoints remained
  when a Deployment was referenced by multiple Services due to premature clearing
  of the terminatedEndpoints map. ([#135172](https://github.com/kubernetes/kubernetes/pull/135172), [@princepereira](https://github.com/princepereira)) [SIG Network and Windows]

ht:calc(1.5em + .5rem + 2px);padding:.25rem .5rem;font-size:.875rem;line-height:1.5;border-radius:.2rem}.form-control-lg{height:calc(1.5em + 1rem + 2px);padding:.5rem 1rem;font-size:1.25rem;line-height:1.5;border-radius:.3rem}select.form-control[multiple],select.form-control[size]{height:auto}textarea.form-control{height:auto}.form-group{margin-bottom:1rem}.form-text{display:block;margin-top:.25rem}.form-row{display:-ms-flexbox;display:flex;-ms-flex-wrap:wrap;flex-wrap:wrap;margin-right:-5px;mar...

## Changelog since v1.3.2

### Other notable changes

* Removing images with multiple tags ([#29316](https://github.com/kubernetes/kubernetes/pull/29316), [@ronnielai](https://github.com/ronnielai))

- Fix Windows kube-proxy (winkernel) issue where stale RemoteEndpoints remained
  when a Deployment was referenced by multiple Services due to premature clearing
  of the terminatedEndpoints map. ([#135171](https://github.com/kubernetes/kubernetes/pull/135171), [@princepereira](https://github.com/princepereira)) [SIG Network and Windows]

- KEP-1880: Users of the new feature to add multiple service CIDR will use by default a dual-write strategy on the new ClusterIP allocators to avoid the problem of possible duplicate IPs allocated to Services when running skewed kube-apiservers using different allocators. They can opt-out of this behavior...