Collector upgrades failing

That is a odd one I haven't seen. The credential field I believe is the encrypted creds/password that the collector uses to communicate with the portal. The creds changes every 24 hours when the collector restarts itself so you don't need to force it. I've had to manually fix it in the past when a collector was rolled back from a DR test or VM snapshot. Was something like that done? You can fix that by copying it from the portal's collector config page to the actual config files on the collector server. But if you are running into that problem, I would expect your collectors to eventually go offline regardless of the upgrade. What version are the collectors now? Are the collectors able to rotate their credentials daily without issue with the current version? Does the cred value in the portal match the creds in the actual collector config files?