A potential use case for this: Currently we (a duo that uses backing tracks) create 3 versions of most backing tracks:
1. The version used in live play (when we used to do that ;-)
2. A version used for member A practice (adding member B's parts)
3. A version used for member B practice (adding member A's parts)
2 and 3 are Smart Copies.
If it had an easy UI (and no latency or mix complications), a simple multitrack capability could eliminate the need for the Smart Copies and separate practice playlists.
It could also be used in live play for situations where some instrument or voice needs to sit out, substituting a pre-recorded version (or vice versa for a guest performer).
To make it easier to manage for these kinds of use cases, I'd prefer some method of marking sub-tracks as 'core', instrument-A, instrument-B, vox-A, etc., and then globally selecting which get played for the practice session or gig. If selections could only be set on an individual song basis, then I think my current Smart Copy approach is just as convenient for my use cases so I probably wouldn't use the multitrack approach.