Softupdates and umount, bug, fixed?

[grarpamp]

At one point in the not too distant past you could...
- newfs -U and mount two disks
- fill one fs with a reasonable hierarchy of a million inodes
- rsync that over to the second fs
- wait till df -i had the same inode count
- blow away some number of inodes from the first, say 5000
- immediately after the rm ; umount the second fs

That would leave you with a difference in df -i count.
And a filesystem that needed fsck -fy
umount didn't seem wait for, or sync, the fs.

Has this been fixed yet?