The big issue about quota support on a globally compressing filesystem is: who owns the one copy of a block of zeros in the filesystem?
In normal filesystems, a block is a block--each user either owns a block, or they don't. In fakefs, a users do not own blocks, the filesystem does. At best, users rent blocks.
Suppose that we have a block B that is shared between user A and user B. Since there are two users, we put half of a block against the quota for each user. Now suppose B deletes their file...B's quota is then transferred to A. If A was at quota before, they are now over quota, even though they did not modify the filesystem!
That might not be as much of a problem as it looks, but it does seem to imply that popular blocks will touch many users' quota records simultaneously. Hmmm.