Every time I have a later release of an album that’s had tracks added, it puts the additional tracks into the correct release and all the original tracks into an earlier release.
IE, the original release had 14 tracks but a later release added 3 “bonus” tracks for 17. So it splits them and I have to drag them to the correct one.
Apparently Picard cannot count. Is there no way to have it count how many tracks are in a cluster and process it like “There are 20 tracks in this cluster. It matches two releases. One with 14 tracks and one with 20. I’ll add them all to the 20 track release.” But nooo, it acts like “Track 1 matches this release, track 2 also does, track n… Hmmm, I’ve hit 14 but there’s still more. Track 15 matches this other version so I’ll put it in that one”. With that way of “thinking” it should go back to the previous tracks and re-assign them to the correct release.
But even after manually re-arranging the tracks, if I scan them again, it screws them up again.
Another frequent issue is Picard not finding the newer release with additional tracks, leaving the ‘extra’ tracks unmatched. But then I’ll search on the barcode or other information and there it is in the database.
Picard uses data from MusicBrainz. Musicbrainz should have separate releases in the same release group for the versions with extra tracks - but if not you can go and add such a release yourself.
You should use Lookup and not Scan for new files. Scan takes an audio fingerprint of each track individually and uses it to identify the recording, and then assigns a release to the track based on the recording. Lookup uses existing information like being in the same directory, numbers of tracks etc. to make a more intelligent guess about the entire album, keeping tracks together.
If Picard still selects the wrong release from the correct release group you can right click on the release and select an alternative release from the same release group.
If Picard selects the wrong release group entirely then you can go and find the correct release on the MusicBrainz web site and then use Picard’s browser integration to load that release into Picard with a single click.
After you have matched the release, Picard saves the Release MBID in the metadata in order to recall the same release next time. If you remove this MBID from the tags, then you will have to reidentify the release next time.
All of this is covered in the (excellent) Picard documentation web site.