The catch is that tilt and accelerometer, while unquestionably useful, are still global. It’s not that we’re expecting these signals in general. It’s that we expect to control them independently on each note.
There are interesting ways that you can marry these signals before output. If you used pressure data to attenuate the range of X axis pitch bend for each note, let’s say, you could bend notes independently of each other, within certain limits. (they’d have to bend in the same direction relative to the starting pitch. it’s not the end of the world – guitar strings have that same limit).
Ditto for Y axis, though I’d handle the attenuation differently. And I guess I should clarify those.
CC74 being in generic parameter space, it might make sense for a released pad to leave that parameter at whatever value was last sent. Pressure is attenuating how much influence a change in tilt has on that value. (I guess you’re adding delta values to a running total for each parameter, so they don’t abruptly jump when a new pad is pressed)
Pitch bend, the pressure should probably attenuate the value itself, so lowered pressure pulls the note closer to center.
I’d probably leave the joystick global.
(The bad news is you’d have to send more tilt data. If six pads are pressed, you’d have to send different X and Y values to each of six channels. But ergonomic limitations are your friend here. It’s not unreasonable to say “this mode is thumbs-only” and limit it to two voice polyphony.)
Anyway, a use case like that becomes a grey area fo me. I’m not sure if I’d call it MPE, but I wouldn’t balk at “Polyphonic Expression”, necessarily.
I’m not convinced that pitch bend is the right use of X for that. Or even notes, for that matter. Most DAW users would be happier with an abstract “this is a handheld 32 knob DJ controller” application. In Eurorack, I’d probably make it a two parameter step sequencer w/ simultaneous control over as many steps as you’d like (because 32 output jacks would be ridiculous).
Point being, “it’s MPE” seems like the wrong claim. it’s treating a rigid standard like a flexible buzzword, when your marketing is better served by “here’s what you can actually do with it” anyway.
(If I owned one, what I described up there is the first application or two that I’d code for it. And I’d be fine with the tilt data coming in globally, because I have the capacity to transform that into whatever I want. But again, this isn’t most users.)
I don’t know. There’s a lot of potential. But is that potential MPE formatted? Probably not.