For what it's worth, I've recently heard some pretty believable
phase-panning software demoed.

One downside is that once you introduce that axis, the image size is
much more noticable; distance "behind" the speakers is also affected by
their separation, so closely-spaced speakers (eg a desktop pair) scale
down the whole "soundstage" and you wind up with a band of deep-voiced
rabbits or something about that size. Admittedly that should be no
weirder than the fact that a TV image is smaller than life-size, but
we're less used to hearing it and it may take a bit of getting used to.