`when: on_failure` jobs only execute once, on the first pipeline failure
For our deployment pipelines we use when: on_failure
jobs for slack notifications. This is a fairly common pattern when there are multiple jobs in a pipeline and you need a notification or some cleanup to happen when any of the upstream jobs fail.
Unfortunately right now, when: on_failure
only executes once, this allows us to receive a notification for the first failure, but subsequent failures on a job retry do not execute the when: on_failure
job.