I only worked on a mixing stage for multichannel once. It looked kinda like this...
The thing is, you can remaster something from the mix downs. You don't have to go into all the elements and completely rebuild the mix from the ground up. With multichannel you have to do that.
I'm afraid I can't help with that because whenever I get a file that is 24/96, I knock it down to redbook then encode to AAC 256 VBR. I don't mess with high bitrate / high sampling rate files on my music server.
The real question, which for some reason people who do tests don't seem to be interested in finding out, is where the line of full transparency lies. At what point does the sound no longer improve with any kind of music being tested? I've done the test myself, so I know, but I don't think a lot of audiophiles do. They think more is better forever into eternity.