Is see that the latter is up to 7 seconds longer, but other than that…?
The version on album My Song sounds the same to me as the version on album Works.
I have a 5:03 recording of Country (rip from tape), but without any metatada, besides the artist name “Keith Jarrett”. Matched by Picard to GB release with this AcoustID (1 source), this is my best bet when it comes to selecting an album to match with.
They should be merged. It takes editors, like you just did, to recognize by data present that a recording shouldn’t be separate. Many recordings, especially within 7 seconds, may be the same due to mastering differences. If you listened to them and they sound identical and since they have the same acoustID, make an edit to merge the two.
I looked at this again (after several weeks of getting more used to editing) and I am actually now in favor of leaving the recordings separated . The one recording appears on “Works”, while the other appears on “My Song”. With this separation, we can see directly which AcoustIDs were submitted to “Works”.
When merged, we would have to look at the “Additional user-submitted metadata” section of all 8 AccoustIDs pages if we were interested to find out which AcoustIDs mostly pertain to “Works”.
But the question is not “do these recordings appear one different releases” but “are those recording the same or different”?
If the 7 seconds difference makes them separate recordings also depends on whether there is relevant audio in this 7 seconds. If one has just longer silence at the end the recording itself would still be the same.