Your model with "the song" being the basis would seem to best fit let's say a musician. The musician might look at his library for inspiration; wanting to quickly access a song as recorded by a number of artists — which album the song came from being of no interest.
No, that's not what I mean at all. In place of "song" think "specific release, mastering, edit"
Take this example..
The Beatles Song "Come Together," original mix, from Abbey Road
Issued on LP in 1969
Issued on CD in 1987
Issued remastered on CD in 2009
Right Now I have
01 - Come Together [Abbey Rd 1969 Vinyl].flac
01 - Come Together [Abbey Rd 1969 Vinyl].mp3
01 - Come Together [1987 CD].flac
01 - Come Together [1987 CD].mp3
01 - Come Together [2009 Remastered].flac
01 - Come Together [2009 Remastered].mp3
all as separate entries in my (imaginary, because I don't use the flacs right now as its inconvenient) library.
It would be much more logical and useful to have
01 - Come Together [Abbey Rd 1969 Vinyl]
==.flac
==.mp3
01 - Come Together [1987 CD]
==.flac
==.mp3
01 - Come Together [2009 Remastered]
==.flac
==.mp3
And then, at the playlist or sync level, where this duplication recurs hundreds or thousands of times over and must be manually managed, I decide the 2009 version is the "go to" I want in my playlists, so I'd just have an entry for 01 - Come Together [2009 Remastered] and select the appropriate version (flac vs mp3, etc) at playback/sync for what I want to do with it.
Playlist setting:
Flac where available, mp3 where not
low space device Sync setting or DLNA setting
mp3 versions only, convert from flac where mp3 does not exist
or
mp3 only, skip those which don't exist in mp3 currently
high capacity device or wired DLNA network
Flac where available, mp3 where not
I know auto-converting is the current (i.e. circa 2005) "solution" for this (across the board, in all programs, this is not a MB specific issue), but that does no good when exporting playlists for use in other applications (sync to an ipod or for Kodi to read, etc). We already spent the time making the duplicate files and have them tagged and arranged, we just want to use them where needed.