Some hypothesis :
* VRAM is not writing pixels at this moment, resulting on a "not newly written line" (badly out of DISP)
* palette offset is badly shifted, palette "masked-line" coming just after data "not newly written line" (but here "masked-line" is about DISP that is to say data... so problem seems here between VRAM and VRAM_palette (to cross with -LittleOne raster problem)) ?
* bad HSYNC interrupt
* bad v_counter range