I have a couple of points on saving...
1. Why not check disk space before performing the save to ensure sufficient disk space, thus avoiding completely running out - I realise it isn't that straightforward because of the yEnc/UUE encoding.
2. Why not block saving only to locations that have run out of space, rather than block all saving? I'm guessing you're processing the save queue sequentially, so this would have to change slightly.
3. When it runs out of disk space it holds the partially written file open, but when you free disk space it starts writing a new file (_2) if the "Duplicate filenames" is set to "auto rename"?! - I would recommend either supporting continuation of the writing process once disk space is freed, otherwise close the file and delete it at the "out of disk space" error.
4. Addition of an option to calculate and record a database of MD5 signatures of downloaded files - this can then be used to avoid saving duplicates. An option to calculate and import MD5 of files already on disk would be handy (including subdirectory recursion). There was a newsreader called "Newsgrabber3" that did this and it was extremely useful to avoid filling disks with dupes.