Edge case when a deploy has multiple running replicasets

[sondr3]

In the screenshot above we have three runnning replicasets, two that are fine and one that is failing. Due to how setting annotations work, when Babylon finds one of the good replicasets and sees that the deployment has a failing annotation it removes it. However, when it then finds the failing deploy it sees that is is missing the annotation and sets it again... and repeat ad finitum. This is most likely a rare edge case and is probably a user error and not Babylon, but it means that the failing application will never be fixed and the team will possibly be notified tons of times.

EDIT: It seems like the above is not the correct edge case, it looks like it is that the failing replicaset takes a while to initialize so that when we loop over all the replicasets for the deploy brukertest-portal-api and the failing replicaset is in the phase PodInitializing it is marked as healthy.