Poly Beebo will do it, it's definitely fun for experimental stuff.
Looks as if I have to consider one of those one day. Right now, there'd be no space on my board left (and it's already the maximum size I allow myself to bring for gigs).
Most modellers are going to have their resources and dev time devoted to the bread and butter stuff though, which is understandable.
I can only partially agree. Computing ressources might still be an issue with this very generation of modelers (even if I doubt it in case of, say, the Axe FX III), but dev time should be almost neglectable. I mean, everybody's doing IR loaders already, all it'd take to make those suitable for spatial FX would be to a) raise the max. IR size and b) offer at least a decay parameter (not being able to limit the length and to fade out would massively reduce usability), pretty much not rocket science at all, almost all DAWs have their own IR plugins these days and there's even some freebies (such as Convology XT).
And as far as computational power goes, I actually don't know whether that's a too limiting factor. When comparing some algorithmic reverbs to Logic's Space Designer, the latter appears to be quite decent in terms of CPU hunger.
And well, unlike with a cab IR block, which needs to run at lowest possible latency, you could as well allow a reverb IR to use up some more CPU cycles - would result in some more predelay on the wet side of things. Heck, many dedicated guitar FX are introducing quite some latency, even if they're made for 100% wet useage (most notably pitch related stuff).
Ok, there's still RAM usage - and I could perfectly understand if larger IRs would exceed what's available within todays modelers. But that's really nothing that couldn't be solved in future iterations. Doesn't have to add much to the cost, either (I mean, even the cheapest mobile phones come with several gigs of RAM these days).