https://issues.redhat.com/browse/OCPBUGS-42106 demonstrated a scenario where the openshift-controller-manager was spinning constantly because of a failure to reconcile. It was triggering client-side rate limiting logic, as indicated by the pod logs, but otherwise not being reported.

    4.18

n/a. https://issues.redhat.com/browse/OCPBUGS-42106 suggests a scenario where duplicate image registry hostnames can be specified in the openshiftcontrollermanager configuration to cause the problem. However, this has been fixed already in 4.18.
As such, it may only be possible to exercise this in a development environment.

    1.
    2.
    3.
    

Client side throttling can persist indefinitely due to a irreconcilable control loop with no direct notification to SRE.

Persistent client side throttling, or the detection of a condition that would cause such throttling, should cause an alert to fire.