After HIVE-2199, the matching logic changed: name-prefix matching no longer works (e.g. GCP lease uses a single character instead of the full pool name).The e2e test still used the old prefix and miscounted; fix is to match by Hive labels instead.

Track by HIVE-3068:

E2E test TestAutoscalingMachinePool in test/e2e/postinstall/machinesets/infra_test.go: creates a busybox Deployment to trigger scale-up, then deletes it and waits for scale-down to the expected worker count.

Observed failures:

1. Busybox pods do not start
   • Error: container has runAsNonRoot and image will run as root.
2.  Scale-down never reaches expected count

• • After deleting busybox, worker count stays at 12 instead of scaling down to the expected 10 (MachineAutoscaler: 2a min=4 max=4, 2b/2c min=3 max=4 → total min 10).
  • Cluster Autoscaler logs show “node group min size reached (current: 4, min: 4)” for nodes; scale-down to (4, 3, 3) does not occur.

Note:
spoke version:  quay.io/openshift-release-dev/ocp-release:4.21.0-x86_64

hive image: "quay.io/redhat-user-workloads/crt-redhat-acm-tenant/hive-operator/hive:hive-on-push-9v9lm-build-image-index"