That's certainly not the only thing that changing inode numbers screws
up. If you were following the NFS thread: the filehandle given to the
client might be kept while the server machine is rebooted. When the
server comes back again, the NFS server daemon searches the entire
volume for the inode (unless there is a more efficient solution). For
synthetic inode numbers, the client suddenly finds itself with an
invalid file handle, or worse, with the wrong file.
I also think some of the GNU utilities make use of the inode
numbers. They are not meaningless, in the way they are often used.
Regards,
Martin