Again, I point to the fact that the "white" stuff isn't white. It's pretty bright, but it has an albedo of about 0.6, far from pure ice. What's happening is that when an impact surfaces some of the bright stuff from below, the thermal segregation model immediately goes to work on the excavated material, and if the situation was such that the native material turned dark in the first place, then the virgin material is going to end up the same way. Maybe it takes years, maybe centuries, but whatever the timescale, it's a blink in geological time. The only bright ray systems we see are very recent ones.

Incidentally, much the same thing is true of ray systems (usually bright on dark) on Mercury and the Moon. Those aren't the only impacts on those worlds -- they're the newer ones. For example, Tycho is estimated to be 108 MYA, Copernicus 800 MYA, and Aristarchus 450 MYA. Those all have salient ray systems. The more typical impact on the Moon, 5 to 10 times those ages, doesn't.

The dynamics of the thermal creep and the hecameter-scale shapes of the boundaries is bound to be an open and interesting question for modelers to address. There are myriad comparable issues pertaining to how winter ice and snow melts on Earth, and I doubt all of them have attracted research interest. (Not sure, even, what FIELD such research would be classified as. Geology? Meteorology? Landscaping? Poetry?)

Climatology, I imagine -- we have studies of glaciation over time, and more recently studies of glacier recession, and of other results of large-scale ice-melt, like the diminution of ice shelves in the Antarctic. However, studies of how this works when there's no liquid phase for the ice to transition through are doubtless much harder to come by -- the most obvious source of information on this subject would be the study of the Martian polar regions.

My first, and continuing, impression of the dichotomy on Iapetus is that the object got splatted with a cosmic tomato. It looks like something was spattered, sputtered or applied to the leading side.

I've found my first example of an impact on the white snowy areas leaving rays from the dark reqolith. N00092009, located in the Voyager Mountains. Looks like an impact dead-centered on a peak.

I did an "RGB" composite image from IR1-GRN-UV3 frames taken at the time. Registration was poor, colr balance was pitiful and I won't inflict it upon you, but the color value of the rays was the same as the weathered regolith surface. An observation presented FWIW.

I don't have a problem believing that in that case we have dark layered on top of white which is in turn (over on the opposite side from the impact site) layered on top of dark again.