David mention this issue here: https://redhat-internal.slack.com/archives/C01CQA76KMX/p1702312628947029

 duplicated_event_patterns I think it's creating a blackout range (events ok during time X) and then checks the time range itself, but doesn't appear to exclude the change in counts?
 

The count of the last event within the allowed range should be subtracted from the first event that is outside of the allowed time range for pathological event test calculation. 

David has a demonstration of the count here: https://github.com/openshift/origin/pull/28456 but to fix you have to invert the testDuplicatedEvents to iterate through the event registry, not the the events