Two jobs (at least) are causing problems in disruption data. We recently fixed a bug preventing most 4.12 jobs from even being scraped for disruption data, it appears a couple jobs are injecting bad data.

From thread: https://github.com/openshift/origin/pull/27430#issuecomment-1255126080

periodic-ci-openshift-release-master-ci-4.12-e2e-aws-sdn-no-capabilities - May be new as of Sep 10, but the job has no console running, which is reporting over 2500s of disruption each run. I don't think we can just skip the test or fix it to pass, as the disruption data would still be gathered and sent to bigquery and showing up in our allowed disruption, anyone querying would need to know to filter it throughout the future. It would also continue to show up in spyglass, which is misleading. Most correct option here I think is to prevent the backend from even starting a monitor somehow.

release-openshift-origin-installer-e2e-aws-disruptive-4.12 - Reporting disruption values of 300s for various backends, normally just a few seconds. This job needs to be omitted entirely, I'd suggest filtering out job names with disruptive in the name.

deads reports there may be code in the importer that assumes 0 if backend data is not present, this would need to be addressed as well.