One man show here. But FINALLY some results.
VEP works like a dream, but for different reasons than expected. I tried one instance of VEP per instrument, and one instance with all instruments loaded inside and routed audio to AUX tracks. Results were the same.
VEP in plugin mode STILL uses the host's audio processing capabilities on the VEP server, but has it's own share of the processor from a Windows perspective (Windows still ran at 30% between the two programs the whole time by the way, it literally does not care what I do inside of Pro Tools
). Thus, loading everything into VEP and changing nothing results in NO performance gains. HOWEVER, they have a handy feature to introduce a processing buffer, and that was more or less the magic fix for me. By setting the size to 4 buffers (roughly a 10th of second), the 88.2kHz session idled at 40% (as before), but never jumped to more than 55% in Pro Tools, even with everything record enabled. No pops, glitches, etc. Roughly that means I'm looking at a .1 second delay before any Pro Tools plugins receive audio, and then the buffer imposed by my interface kicks in.
Bear in mind my workflow involves virtual instruments exclusively and since I sequence and edit by hand, requires no live input. So the total delay literally makes zero impact on me whatsoever.
So, now I've found myself two good workarounds that have zero negative impact on audio quality, but the original problem is still unsolved. How can I increase my capacity for real-time processing? I'm happy to solve it with a software shim in the meantime, but why do I have to do that at all with a 30% CPU load?