Image registry has a known limitation where only 2 replicas are supported. During the test for

[sig-node] NoExecuteTaintManager Multiple Pods [Serial] only evicts pods without tolerations from tainted nodes

if both image registry deployments are on the two tainted nodes this will lead to increase in disruption.

We discussed options to handle this case where it is known to be caused by a specific test. One consideration is to filter out the disruption that overlaps with the test run in EvaluateTestsFromConstructedIntervals another proposal was to disable image-registry disruption for the serial job or potentially flake the test.