Component readiness failures on: 
 [sig-autoscaling] [Feature:HPA] Horizontal pod autoscaling (scale 
resource: CPU) CustomResourceDefinition Should scale with a CRD 
targetRef [Suite:openshift/conformance/parallel] [Suite:k8s] 
            
            
Always looks like: 
                {  fail [k8s.io/kubernetes/test/e2e/autoscaling/horizontal_pod_autoscaling.go:211]: timeout waiting 15m0s for 2 replicas: Told to stop trying after 60.251s.
Unexpected final error while getting int: Operation cannot be fulfilled on testcrds.autoscalinge2e.example.com "foo-crd": the object has been modified; please apply your changes to the latest version and try again
At one point, however, the function did return successfully.
Yet, Eventually failed because the matcher was not satisfied:
Expected
    <int>: 1
to equal
    <int>: 2}

    HPA OpenShift 4.20

    ~5% reproduce rate, but it does flake 

    1.Run that test case in a loop
    2.Observe failure when the status update to the CRD happens _after_ the HPA updates the custom resource 
    3.Profit 
    

    Test case occasionally flakes

    Test case does not flake

    This is such a strange failure because the failure happens as part of a weird hack upstream is using to imitate controller behavior -- it updates the status with the replica count (the HPA needs that info to scale) but sometimes the HPA is too fast and it loses the race, and the retry function doesn't currently regard a collision as a retryable error so it just fails.