Terminating pod logic is error prone
Both of these are error prone. Essentially the issue happens when the initial check finds some pods in a terminating state but they terminate before it hits the kubectl wait.
We need to better account for this situation.
Example failing pipeline: https://repo1.dso.mil/platform-one/big-bang/apps/core/cluster-auditor/-/jobs/8081354
Edited by Micah Nagel