When a fix is integrated in multiple repos at the same time (typically open and Oracle closed), the Skara notifier bots seem to often end up in a race for creating a suitable backport record, as separate threads are processing each repo concurrently. While this isn't causing any data loss or corruption, we currently get a SEVERE notification each time it happens. I think we need to more actively handle this situation more gracefully to avoid notifying admins when there really is no problem.

I think a reasonable strategy would be to have a fallback when backport creation fails, to check again if it already exists, and if so get and return the existing backport instead.