I scanned some Tri-X negatives developed in Caffenol as color (back in 2003) and got some realistic looking color images, more or less like two-color Technicolor. I doubt there was any actual color information (since Tri-X has no dye couplers and uniform panchro sensitization), it was mostly differential levels of stain doing it -- but the result looked remarkably good. I'm at work now, so can't attach a scan, but I'll try to remember to do so when I'm home...