Software path-tracing has been on my bucket list, mostly to test a concept: physically based instant radiosity. If an eye-ray goes camera, A, B, C, then the light C->B forms an anisotropic point source. The material at B scatters light from C directly into onscreen geometry. This allows cheating akin to photon mapping, where you assume nearby pixels are also visible to B. Low-frequency lighting should look decent at much less than one sample per pixel.
Software path-tracing has been on my bucket list, mostly to test a concept: physically based instant radiosity. If an eye-ray goes camera, A, B, C, then the light C->B forms an anisotropic point source. The material at B scatters light from C directly into onscreen geometry. This allows cheating akin to photon mapping, where you assume nearby pixels are also visible to B. Low-frequency lighting should look decent at much less than one sample per pixel.