openkruise/kruise v1.9.3 on GitHub

What's Changed

WorkloadSpread: trigger reconciliation when a Pod becomes unschedulable — #2634 by Ishaan400 (@Ishaan400).
Ensures the adaptive scheduling strategy reevaluates pending Pods after the configured rescheduling timeout, rather than waiting for an unrelated event.
WorkloadSpread / controllerfinder: prevent an empty scale selector from matching unrelated Pods — #2642 by sarankumar (@sarankumar1325).
Treats an empty selector from a custom workload's scale subresource as matching no Pods, preventing WorkloadSpread from claiming every Pod in the namespace. Selector conversion errors are now returned instead of ignored.
CloneSet: clear stale rollout timeout conditions on a new update revision — #2600 by Abdullah Zulfiqar (@AbdullahZulfiqar2005).
Removes the previous rollout's ProgressDeadlineExceeded condition so a new rollout can start with a fresh progress deadline. The backport retains the existing v1alpha1 API.
BroadcastJob: exclude failed Pod creations from active counts — #2620 by Sarthak Sharma (@sublimemonk).
Increments the active counter only when Pod creation succeeds, avoiding inflated status.active after creation failures.

StatefulSet tests: wait for PVC and StorageClass informer caches to sync — #2635 by Zhiting Zhu (@photoszzt).
Prevents flaky test results caused by reading these caches before they are populated.

Full Changelog: v1.9.2...v1.9.3