But you do: That’s what ARs do.
My point is that the person doing the tagging knows what they want to do, and can script accordingly. Not that the example will work perfectly for what everyone wants to do. Maybe someone wants "$set(artist,$if2(%performer%,%composer%,)). Maybe it’s "$set(artist, %performer[orchestra]%).
It doesn’t really matter what precisely someone wants, they should be able to script it into whatever they want without changing the track/release artists in Musicbrainz proper. It doesn’t even matter whether Picard is currently capable of getting the exact effect someone wants (e.g. there is nothing like “$recordingartist” in Picard). The principle is that the way someone wants to use the data is independent of what guidelines should say about how Musicbrainz itself wants to display that data.
Picard can always be improved; other classical-specific tagger utilities can be developed, but the data is all there in the ARs if you want to use it.