Wait for backing resource teardown before removing ClusterInstance finalizer

[caxu-rh]

Problem

deleteIfExists treats an accepted delete request as completed teardown. The ClusterInstance controller can then remove its finalizer while backing resources with their own finalizers still exist. The pool reuses deterministic instance names, so a replacement can encounter the old HostedCluster, VM, or disk while deletion is still in progress.

This can cause stale readiness, name conflicts, or incorrect resource reuse between lease cycles.

References

• internal/controller/helpers.go:42-54
• internal/controller/clusterinstance_controller.go:613-627
• internal/controller/clusterinstance_hypershift.go:411-435
• internal/controller/clusterpool_controller.go:531-547

Suggested direction

Make teardown distinguish complete deletion from a delete request that is still pending. Keep the ClusterInstance finalizer and requeue until backing objects whose names will be reused are absent.

Tests

Add a backing object with a finalizer and verify that the ClusterInstance remains until the object is gone.