we had the same problem and we have used something like 4 tickets to the parallels support paid system to try to solve.
we have listen to a series of answer like "increase the bandwith from the plesk server to the backup ftp server" (done: if the trasfer time of the single ftp backup task is more than an hour and an half "real", 2 hours "in theory", the task crashes and the backup remain appended) , "use active FTP" (done, changes nothing) "use passive FTP" (done, changes nothing) to reach to the final answer "bugfix comes in the next version". at that time we were using 9.2.1 release and the 9.2.2 are in deployment. apply the 9.2.2 have changed nothing. in the 9.2.3 changelog we haven't see nothing about ftp backup, but we havent' tried again to see if something changed. and we haven't a great desire to spend much more money to the support system to haven't a solution over that problem.
what we have done is a couple of Crontab script (one over the Plesk Server and one over the FTP server) that checks the backup files and delete the ones in excess.
THAT mitigates the problem for us. don't solve, but at least we don't go again out of disk space or we don't see anymore customers domains blocked from the quota system without a reason.
But our script is based on backup file names, that we have standardized, and directory where the files are stored over the FTP server, so i don't think that it can be useful for someone else that have a different infrastructure than ours.
what i can say to you is to make the same thing: if you want to use the FTP backup system , consider to create those shell scripts for your environment