Then: Someone has to initially send such an AcoustID (or more exactly: a fingerprint) for a specific song with some metadata to the AcoustID-Server. From this moment, many other users can make a lookup with the same AcoustID and get back the previously submitted metadata.
If someone would transmit WRONG metadata for a song, it could happen that you get this wrong data back for your song. There are several threads how you can fix such situations.
Normally if you load files to Picard they will end up on the left pane as unmatched files. You then have to use either “Scan” (will use AcoustID fingerprints to identify the songs) or “Lookup” (will search Musicbrainz.org based on existing tags) to identify it.
If the files already go directly to the right pane and are automatically matched to albums that means the files had already stored the MusicBrainz IDs. And if those are wrong it means somebody has previously tagged those files against the wrong tracks.
That’s only half the truth. If the “Automatically scan all new files” option in Options → General is checked, files in the left pane will automatically be scanned and end up in the right one, just like using “Scan” manually.
@Mike_Rankin I have seen bad AcoustIDs appear for a few reasons. Usually always caused by a human mislabelling something. Often a large compilation getting uploaded, or a release with a misprinted cover, or a dodgy bootleg with a bad track list. Or just a sloppy user who didn’t know what effect the Submit AcoustID button did…
Anyway, it is easy to correct. You can follow the AcoustID data through to the AcoustID website and errors like you describe often stand out quite well and can be corrected.
There are better experts here who can talk about how it all works. But this is a good example of why any computer matching always needs a bit of a check over.