@JSWolf: FWIW, I don't have access to one, but Mk. 6 devices *may* be able to do things differently, as another data set

.
(Depends on if the Mk. 6 kernel code matches the Mk. 5 nickel code, which I obviously can't say ^^. What I can say is that I hadn't seen anyone else implement it like that in kernel. Amazon tried some different stuff, but it was mostly aimed towards low-bitdepth experiments. All of this is testable w/ the latest FBInk release, behind the legacy dithering mode, for curious parties with shell access

).