I could see it within the realm of possibility to have the ability to do an "SSS hold". Meaning as long as you keep switching between Performances that support SSS - that the currently sustained notes would not drop off. That prioritizes the "oldest" held notes over the new ones. So A->B the sounds from "A" would SSS hold. Keep holding and switching from B->A even though you may have held notes from "B" - they would be cut off (not SSS as they normally would) and instead the held notes from "A" continue.

Likewise even the following should be possible.

A->B (A held notes kept with SSS) B->C (A held notes kept with "SSS Hold", any held notes from B are cut off).

I say possible in that it would be possible, I am speculating, to implement.

What I would do - and what I primarily do for all of my work now - is construct a single Performance that has everything from what you're saying is "Performance A" and everything from "Performance B" as a single Performance. Then use keyboard control, mutes, keyboard range, velocity switching, XA control, or a number of other techniques to keep everything within the context of a single Performance without the need to switch away to a second performance. Then I could go back/forth at will without issue or limitations.

I play medleys with several songs that each have their own sounds. I haven't run out yet using 8 PARTs. Generally what I do is slightly compromise and reuse sounds between these songs within the medley. Strings, leads, etc. I carefully plan where on the keyboard everything is laid out and use the SCENE buttons to automate bringing in/out different PARTs if I need to remove some part of a layered sound or otherwise share the same keyboard range for two sounds I do not want layered.