/vr/ - Retro Games

MANY reasons.

First of all, Megaman's use of the technique is part of the reason those games are notorious for their flickering and slowdown. Keep in mind that:
-The NES's processor is so slow that even programming at a machine language level is rarely enough to process more than a few characters onscreen at once, especially if you also have to process AI, animations and the music engine every frame.
-There is an eight-sprite-per-scanline limit. This horrible limitation is what causes flickering (keep in mind, sprites on the NES are limited to either 8*8 or 8*16 pixels, so making big characters requires moving multiple sprites in unison). Layering sprites for better color depth further eats away at this limitation - the more you do it, the more susceptible to flickering you are.

Second, remember that graphics on the NES are not stored in a recognizable format like PNG or GIF. They're essentially just raw indexed bytes that you load into VRAM and assign a palette to. The NES only lets you use four background palettes and four sprite palettes. (Pic sort of related: it's two of the sprite palettes used in SMB1. I couldn't find a better picture with all four). This means you have to use color conservatively: if you use two palettes to create one detailed layered sprite, you can, but then you only have two more palettes for every other object on screen. This is why Megaman only really uses it to add detail to his face - the white and flesh tones are ambiguous enough that they can be reused in another sprite.