Ok, we’re getting somewhere!
That makes much more sense - IRs are capable of producing nonlinear responses, but the nonlinearity itself is baked in, right? We’re in agreement there?
I’d continue, then, that if maybe theoretically that isn’t 100% irrelevant, in practice, for our purposes, it’s something like 99.99% irrelevant. While the nature of that nonlinearity itself is volume sensitive… one, you can pretty much throw out the very extremes becuse you want SOME speaker involvement so any modeling below a threshold where that happens doesn’t matter, and two, for at least most genres and styles (doing “Rumble” covers is an exception, I guess), you want it to stop well before the point of "total speaker meltdown.
Basically, there’s a sweet spot you probably want to fall in. You can debate how large it is and where on that spot you want to be, I guess, but safe to say it’s a range, but not an unbounded one. And, within that sweet spot, from the speaker’s point of view, the dynamics of the performance I’d argue, at least for our purposes here, are pretty minimal; the difference between gently plucking an open high E, and banging on an open E chord Townsend style, into the red channel of a Rectifier, are going to be mcuh, much, much smaller than moving the master volume from 3 to 4.
That’s not true across the board, sure - but, it’s much more true for “Enemies of Reality” than “Riviera Paradise,” ironically enough.
tl;dr - it’s pretty safe to assume that for a hard rock/metal player and a contemporary high gain amp, the degree of nonlinearity in speaker response coming from performance and not how loud the amp is assumed to be cranked up is pretty negligible… and if you’re talking about testing how “accurate” a model is by things like null tests, speaker nonlinearity is going to have a vanishingly small impact compared to things like whether the SM57 is a millimeter further back or two the left on the real cab than the IR, how worn in the speakers are, condition of the tubes, etc. And, if you want to test that faithfully, you have to either run both through a real cabinet, or both through an IR, and for that reason you can’t really do a truly scientific comparison of a real amp to a modeler/IR since you can’t really fully isolate one from the other.
Yes, I have heard of people making “room sound” IRs. But, they’re very rare, and the broader point here is you can’t compare an amp to a modeled version unless you’re also running both though either the same IR, or the same 4x12 cabinet in the same room, because of how much even tiny changes in mic position change the sound, much less going from a close mic to listening half a dozen feet away from the speaker.