When using Hive to create OpenShift clusters with autoscaling enabled, after adding workloads, the scale-down functionality fails to operate correctly, resulting in the cluster being unable to automatically scale down based on actual workload requirements.    

quay.io/openshift-release-dev/ocp-release:4.21.0-x86_64 - Failed
registry.ci.openshift.org/ocp/release:4.22.0-0.nightly-2026-01-31-082403 - Failed
quay.io/openshift-release-dev/ocp-release:4.20.13-x86_64 - PASS!    

    Always

Steps to Reproduce:
1. Create a Hive ClusterDeployment with a MachinePool that has autoscaling (minReplicas=10, maxReplicas=12) on a multi-AZ platform (e.g. AWS 3 AZs).
2. Wait for cluster to install and scale to min 10 workers; MachineSets stabilize at [4, 3, 3].
3. Deploy a workload on spoke cluster (e.g. a busybox deployment) that consumes capacity so the autoscaler scales up to 12 workers; MachineSets become [4, 4, 4].
4. Delete the busybox deployment. 
5.The autoscaler should scale the pool back down to min 10 workers.

cluster-autoscaler logs:
I0203 23:01:16.484087       1 eligibility.go:163] Node ip-10-0-46-119.us-east-2.compute.internal unremovable: memory requested (52.8758% of allocatable) is above the scale-down utilization threshold

    Scale-down to 10 never occurs; worker count remains 12. 

Cluster Autoscaler should consider scale-down for all worker nodes in all three MachineSets.
Two of the three MachineSets have min=3 (pool min 10 distributed as 4+3+3); 
CA should be able to scale down one node in each of those two MachineSets, reducing total workers from 12 to 10.