This sort of strategy feels to me* like it would be ideal in a gamedev sort of environment - considering the nontrivial practical distance between each of the layers in this scenario, I can totally see each layer occasionally showing up out of alignment, or rendered in the wrong font, etc. My (very abstract and non-practical) impression is that the rendering pipeline in a game is typically much tighter/cohesive than an entire desktop shell.

* I could be wrong; I'm just spouting naive hot air with no practical experience at either end of this spectrum :)