When checking meta-info already available for a Chopin release (https://musicbrainz.org/release/e08256f5-8783-437d-9493-cbc7acbf8c4b) against a CD box I own, I found that some records in the MusicBrainz DB have 2 links to the same performer and 2 links to the same work with different recording dates, for example https://musicbrainz.org/recording/e16a5f0c-eaad-4e3f-8339-2bbdc9445f31. This looked wrong to me, so I investigated a bit further.
It appears that affected records appears on different releases, where different information about recording date and place is given. For example, the first record appears on release https://musicbrainz.org/release/b1fd0d5e-d35b-4f82-8013-3696327d3542 and https://musicbrainz.org/release/6e3cb2ae-1ea1-43df-9e21-47e902a89d5c. The first release points on a discogs entry stating that the recording was done in London in December 1981. The second release points on a discogs entry stating the the recording was done in Germany in April 1982. The CD box I own states that the recording was done in Germany in 1982 (without mentioning the month), which agrees with the second discogs release. Theoretically it is possible that these are two different recordings done in different places on different times, but duration of all affected tracks is absolutely identical, so the probability of such explanation is quite low. I am inclined to believe that one of the releases on discogs gives wrong recording place and time, although I can’t prove it.
What is the right approach in such case? Currently MusicBrainz DB links a record to the same work and performer twice with different dates, but somehow it looks wrong to me. Maybe it would be more appropriate to keep only links with dates mentioned by majority of sources, and add an annotation that one source gives a different date?