I noticed a lot of discussions are very PPU clock cycle specific (loading HT registers on CC 252 etc.), however if I'm not mistaken is such timing even feasible?
The problem is since you don't know which instructions are being executed before hand, the cpu may in fact not have ran for 84 cycles (252 / 3, again the HT counter thing), it may have ran for 86, or even 87. Each scanline could be at a different cpu cycle at this point, and every frame may be different also for a given scanline.
Just wondering if this is a problem at all. :) I can't see how people can get such super specific pixel rendering timing, when in fact the cpu's cycles may be different then what they were shooting for.