Just submitted a couple of N64 fixes via pull request.
8bpp mode for framebuffers was completely nonexistent. After some back-and-forth with some N64 test programs courtesy of krom and my 64drive, it matches a real N64's output almost exactly:
There's some difference in gamma, but beyond that, it seems to work. Bit interesting in that the VI still scans the framebuffer out as 16bpp 5:5:5:1 pixels, but the neighboring 8-bit values that exist in the framebuffer effectively get paired together. I suspect if the test drew fill-rects with an odd number of pixels rather than even, we'd see a fringe of "wrong" colors on one or the other side of each column.
It also turns out that the "Magic Matrix" used for some dither operations was transposed along its axis. Fixing that fixes the dither patterns used in krom's AlphaCompare test: