That means if you notice a recording which exists twice (for example in different releases) with the same attributes (name, artist, length and probably ISRC), you should merge them:
That makes it sound like they should be merged but then the style guide for the recordings page has the following blurb (Style / Recording - MusicBrainz):
Audio recordings of the same performance from different sources will always have different audio. A new MusicBrainz Recording should be created for each track created from different audio recordings.
That’s a little confusing for me.
I understand having separate recordings if one is from a studio album and one is from a live performance, for example, but if it’s the same track from different studio releases, shouldn’t it have the same recording?
“Same performance” refers to live recordings.
As example, the bootleg from John Smith is different than the bootleg from Little Jimmy even though they were both at the Greatfool Did concert on February 29 1969 at George’s Dome.
Basically, same recordings are the same recordings. Just because it is the same song by the same band doesn’t mean it is the same recording. Just because it is on two different albums does not mean it is different recordings.
Without listening to the tracks, I would assume your example is probably the same recording.
One recent poster on the forums is now looking at whether AcousticIDs can be used to create matches such that AcousticID of Recording A = AcousticID of (Recording B - xmilliseconds).
Not even sure if it is possible. But it would be a great help in merging Recordings.
In your case it is almost certainly the same recording. Usually a studio recording same artist and time is pretty much good enough. As others have said the stuff you are referring to regards different people recording the same live performance.