I'm also one of those who sometimes hears soundstage differences between DACs/amps (even when they are considered "blameless" -- fully transparent -- wrt the usual standard measurements). Of course when precision level-matched electrically (which btw never is perfectly possible as soon as you have the tiniest frequency response and/or distortion differences for reasons I think I don't need to explain as it is immediately obvious. And it's interesting that these sound stage differences are sort of immune to tiny level mismatches below 0.3 dB or so). And properly blinded. This is really puzzling my engineering and scientific self, believe me.
It is obvious that there must be a signal difference which must show up, for example, in detailed subtractive analysis (which again is an art and a science in itself, I've spent more than a decade on that topic alone). I've done my own private research on that, with no conclusive results as of yet. I only have some pet theories, one of them is microscopic "time smear" that already is buried in the noise floor, technically, but maybe still relevant perception-wise (as we can hear quite a bit into the noise floor).
Stereo illusion from two-channel content played back with two speakers is very fragile, and soundstage depth/width, size of phantom sources etc is especially fragile and, as noted, mostly a thing of mental construction. And that might be just the point, very subtle things that make it harder for the brain to construct a soundstage illusion from non-realworld-like cues, differing between DACs/amps. The soundfields near the ears creating the phantom localization is never like what you'd have from real sound sources, it's artificial, and percieved soundstage is an aquired skill by training.
My 2ct. and strictly IMHO.