I have observed this in the past, but had assumed it was fixed by the changes we made to the finalizer removal. It seems to still happen though. Note that this just looks bad in the logs but has no other effects; the operator restarts and will happily build and destroy SMCPs afterwards.

Method to reproduce:

• create an SMCP resource, including jaeger&kiali (this seems to be important because it slows down pruning)
• wait for it to come up
• delete it
• as soon as kubectl delete returns, kubectl apply it again

-> the operator will panic trying to remove the finalizer from an already deleted object, possibly even from the new SMCP resource
-> it will immediately restart and continue working
See logs at: http://pastebin.test.redhat.com/788687

This seems to be related to the retry-loop for the finalizer removal; we should possibly get rid of that.