Can we all take a little step back and think about the implications of the original patch from Alex? Because I think there is very little. And updated sunit/swidth is just a little performance optimization, and anyone who really cares about changing that after the fact can trivially add those to fstab. So I think something like his original patch plus a message during mount that the new values are not persisted should be perfectly fine. We can still make xfs_repair smarter about guessing the root inode of course.
