I was on that panel, and yes, several of us talked about stems, but nobody said they wanted them in order to remix the record or change the mix, And everyone agreed that would be an unwieldy, time-consuming, costly, and probably inappropriate thing to do. IIRC, it was stated: if the definition of mastering is to put the finishing touches on a record, than anything that helps to achieve that goal is potentially useful. The use of stems for mastering, I guess, is borne, not from a desire by mastering engineers to screw around with the mixes, but by the frustration sometimes shared by MEs and their clients of not being able to fully optimize the end result when there is just the stereo mix to work with. If it's easy to bring in stems and - for example - separately EQ the bass stem to achieve the desired results, than why not?
If you don't want or need to make those kind of changes in mastering, than it's simple: keep on delivering a stereo mix. If you want that kind of flexibility and have the budget to book longer mastering sessions, and the discipline to not go nuts with this, than think about stems. There are many paths, grasshopper.
I've only mastered stereo projects from stems a few times, and everyone involved was very happy with the process and the results. I've also mastered a couple surround projects from stems; working in surround it's great to have those options since there are so many variables and unknowns in that realm.