Klaus,
The original service pack update you had wasn't necessarily optimally delivered. Because your large configurations were not tested for a week after the update, you were committed to that Service Pack. Whereas if that were detected during upgrade testing, the rollback would not have been multistep.
The need for rollback would have been avoided by dropping the Undo disks and rebooting - one step. Instead to go from a newer version to an older version did require some extra steps. These are public forums so I can't comment on future enhancements we may add regarding automatic upgrades, we are aware this is a possible area of improvement as we are our largest customer too.
Having a backup of the infrastructure VMs helped, and if you remember, the issue with the large configurations still happened after the rollback, hence I left a hot fix in place for you. I can't explain why you didn't see the problem before the upgrade, it could have been an environmental change that occurred during the week between the Service Pack being applied and the usage of the large configurations. I don't know.
FWIW, I still haven't reproduced your issue in-house, but I do have contingency for it in the upcoming Service Pack so you may avoid any problems.. It's an interesting issue, btw, and I would be happy to provide your usage scenario to Microsoft for further analysis, if I can reproduce it.