Description of problem:

The integration test "should delete the ServiceMonitor when mgh deleted" was flaky with a 75% pass rate. The test failed intermittently due to a race condition caused by the two-phase cleanup behavior in the manager reconciler's pruneResources function.

Version-Release number of selected component (if applicable):

Global Hub 1.7.0

How reproducible:

Intermittent - approximately 75% pass rate before fix

Steps to Reproduce:

1. Run the integration test "should delete the ServiceMonitor when mgh deleted"
2. The test calls reconcile() once after deleting the MulticlusterGlobalHub instance
3. Observe that ServiceMonitor deletion may not occur in the first reconciliation

Actual results:

The test failed intermittently because:
- Phase 1: pruneResources deletes ManagedClusterMigrations and returns early
- Phase 2: On next reconciliation (automatic rescheduling), ServiceMonitor gets deleted
- The test only called reconcile() once and relied on controller's automatic rescheduling, creating a race condition

Expected results:

ServiceMonitor should be deleted in a single reconciliation call, eliminating the need for multiple reconciliation cycles.

Additional info:

• Root cause: Early return in pruneResources prevented ServiceMonitor cleanup when migrations existed
• Fix: Modified pruneResources to delete both migrations and ServiceMonitor in a single reconciliation call (PR #2131)
• Impact: Integration test flakiness and inefficient resource cleanup (two reconciliations instead of one)
• Test results after fix: 4/4 test runs passed (100% success rate)

🤖 Generated with Claude Code