Really interesting discussion, but a question arises: could the method to capture IRs hide some phenomena that occurs only on some circumstances?
I mean, an IR is almost always captured with a sine sweep which is just a single frequency being reproduced by the speaker at a time, but this would not detect for example some irregular motion of the cone that might occur when the speaker tries to reproduce more complex sounds.
Could be interesting to compare IRs of the same cab+mic configuration captured using different test tones (white noise, pink noise, double sine sweep, music program) to see if the frequency response remains the same across all of them.
But maybe someone has already done this kind of tests or maybe my knowledge is too limited to talk about these things and my argument is just "bovine stools"