Revisions allow you to track differences between multiple versions of your content, and revert back to older versions.
This is a critical bug, preventing certain multiple group threshold models from working correctly. It may be invisible to users and they may be getting incorrect results but unaware of this fact. It does seem more cosmetic than functional; the minimum and the -2lnL do not agree. Parameter estimates may also be "stale" at the end of the run. Can we up the priority on this one?