Configure topology spread constraints with minDomains and nodeAffinityPolicy for zonal autoscaling

Steps

1. Add a topologySpreadConstraints entry to the pod template with topologyKey: topology.kubernetes.io/zone, whenUnsatisfiable: DoNotSchedule, and a labelSelector matching the workload's pods.
2. Set minDomains to the minimum number of zones the scheduler should treat as eligible even if fewer currently have ready nodes (e.g. minDomains: 3 for a 3-AZ cluster); this only takes effect when whenUnsatisfiable is DoNotSchedule, and is stable/GA since Kubernetes 1.30 (beta-by-default since 1.27).
3. Set nodeAffinityPolicy: Honor (the default) so skew is calculated only over nodes matching the Pod's own nodeAffinity/nodeSelector, keeping zonal skew accurate when a cluster-autoscaler node group spans zones unevenly; nodeAffinityPolicy graduated to GA in Kubernetes 1.33 (beta-by-default since 1.26).
4. Optionally add nodeTaintsPolicy: Honor so newly added but still-tainted autoscaler nodes aren't counted toward domain eligibility until they become schedulable.
5. Apply the workload and verify placement with kubectl get pods -o wide --sort-by=.spec.nodeName, and confirm every node carries the topology.kubernetes.io/zone label with kubectl get nodes -L topology.kubernetes.io/zone, since minDomains and nodeAffinityPolicy calculations depend on that label being set correctly.