CI payload runs have failed 3 times with disruption failures. But a close look reveals that every time the disruption is bad, there are more than 3 masters showing up in the spyglass chart. 

kube-api-new-connections mean disruption should be less than historical plus five standard deviations

openshift-api-new-connections mean disruption should be less than historical plus five standard deviations
 

Take this job as an example. There are four masters in the spyglass chart. Yet not all those four are showing under nodes gathered by gather-extra.

We are going to attempt a revert to see if that might be the cause.