Fake Koola’s sample scene also doesn’t seem to propagate emissive lighting using an inverse square falloff rule, but i may be wrong about that.
In any case that’s a pretty impressive tech if you think about the constraints in which it was developed (one and a half man / native engine integration / spare time dev / no resources), so I still have my hopes up.
I would be pretty intrigued to see same scene comparisons with AHR and experimental DFGI in 4.7!