Recordings with multiple AcoustIDs
This recording has four different AcoustIDs (look at the Fingerprints tab). This is almost certainly due to the huge variety of different releases the recording is listed as belonging to: the slightly different versions each appear on a different subset of the releases listed.
For that recording in particular: The hard part is determining which tracks belong to which AcoustID; the only real way to do it is to acquire copies of the files which are definitively from each of those releases, find a recording which matches only that AcoustID, and move the track to that recording.
Reading an AcoustID page
Take a look at this one
The first thing you see is a list of fingerprints. These are the slightly-different audio selections that were considered similar enough to get the same AcoustID. From there you can select a couple of the fingerprints to compare how different a “same” fingerprint might be. 11831566 and 11185730 are the most-submitted ones, so let’s take a look at those. Open each in a new tab so you can easily switch between them.
The vertical axis is time. The whole bar represents 2 minutes (or 1 minute in a few cases, due to the older fingerprints), so each vertical pixel is a small fragment of time.
Once you have both those fingerprints open, you can see that the difference between them is very small: only a slight vertical shift, plus a single pixel different here and there.
When not to merge
Compare it to either of the other ones and you can see that it is substantially different. Similar enough that it is probably the same song (the general appearance is the same), but still different enough to be almost certainly a different recording in some way. Of course, there’s no way to tell exactly what the differences are without hearing it.
When to merge
Finally, lets look at the one which is a different acoustID but I still think would be the same recording: The most-submitted fingerprint is 21014712. Comparing this back to 11831566 you can see again that they are practically the same. IMO recordings having these two acoustIDs are probably OK to merge.
Further down on the AcoustID page you can see a list of recordings which have this AcoustID. In general, in my experience it is safe to merge recordings which have the same AcoustID, as long as they each only have the one AcoustID. If they each have multiple AcoustIDs its probably because some releases have a different recording and someone picked wrong. Also be mindful of different performance ARs and ISRCs (though liner notes and the various ISRC sources have been known to be wrong as well.)
There are a few places where AcoustID has trouble.
False matches: Karaoke versions; instrumental versions; radio edits (where only a small amount is edited, bleeped out, etc.). Recordings which only diverge after the 2-minute mark where the acoustID fingerprint ends. I have been told that very short (15–30s) tracks also have this problem, but I haven’t seen it myself.
False differences: A time-shift of more than a few seconds will often cause a new/different AcoustID to be assigned.
It is also worth noting that recordings that differ by more than 10 seconds will always be given a different AcoustID, even if the fingerprint data matches 100%. So if you see an acoustID with a track attached that with a difference greater than that, it always indicates something wrong (I’ve seen the recording being simply the wrong duration, and also recordings from two releases that shouldn’t have used the same recording.)