The vblank timing could cause real problems.
aside from that, there might be one or two that require sprite timing. The gross inaccuracies in the NES are already gone, though
SNES is disorganized as ever, but the BRR bit comes from anomie's documents that are included with "vSNES" (save state viewer for most emulators that support them).
The H-DMA behavior is byuu's news, but as far as the PPU dot-by-dot behavior, I don't think that is actually recorded anywhere. The offset-per-tile oddity is well-known. The dot-by-dot just explains why Hook fails to glitch on real hardware
(for the interested, the SNES has 16-bit VRAM access, and has to cache two tiles at a time. This likely happens without regard to what BGs are active)
Probably the most important discovery lately is that H-DMA on the same channel as DMA terminates the DMA on the first H-DMA occurance.