The way I see it - if that is your goal you only have two options:

• developing your own protocol, drivers for each platform and user friendly software to utilize it - this implies high support cost and limits 'mass storage' to only supported platforms, or
• building a shadow virtual FAT on top and exposing it to USB - it's not as difficult as it sounds, I did some rudimentary work in that field a while ago and I'd gladly donate you the code if only I could find it, however I'm not sure if Microsoft still claims patents on that so it might get costly

Even Apple wouldn't dare to break legacy USB-FAT so the latter is pretty much Apple-proof, and not only that - every device with that support would be able to read the storage.

If we move away from legacy, as I've mentioned earlier, current Linux filesystems wouldn't need much modification to do away with single-host mounting, but that doesn't solve the problem at hand.