TestDRA/all/ControllerManagerMetrics still waits a fixed 200ms before
checking the success counters. In CI this can fail with:
- Max difference between 1 and 0 allowed is 0.1
- success metric with admin_access=false should increment
The controller work then continues into cleanup, which produces
follow-on context canceled and handler timeout noise.
Replace the fixed sleeps with polling on the metric values so the test
waits for the controller's counters to converge.