Picard doesn't seem to detect duplicated or malformed tags. If a file is loaded that has duplicate or malformed tags, such as partial MusicBrainz information, Picard will fail to lookup correctly. If I then do a manual lookup, the new data will be added, resulting duplicates to the files.

IMHO duplicated tags should be fixed before a lookup request is made.

EXAMPLE:
I've got a master music library that's mostly in FLAC on my external hard drives, so I use SoundConverter[1] to make MP3s for my various devices. There's a bug[2] in that application which causes it to not write MusicBrainz tags correctly, but a work-around would be to have Picard detect potential duplicates and [offer to] fix them before attempting a lookup. (There are probably other applications which munge files like this.)

The first screenshot shows Picard operating on a file that's been converted. The highlighted fields are the ones the converter gets wrong, but there are more that I didn't select, as can be seen in the second screenshot.

[1] http://soundconverter.org/
[2] https://bugs.launchpad.net/soundconverter/+bug/1174949