It's difficult and memory-expensive to do subpixel font rendering in the presence of things like transparency/colors/blur effects - eventually you'd end up in a situation where it's just missing half the time anyway.
How expensive would it be? Windows XP (circa 2001) does it without GPU acceleration with ease on a 512MB ram machine. Windows Vista runs on 2G of ram, with transparency everywhere (to the point that it is distracting).
A lot worse than that. They didn't have 5K HDR 120fps displays and macOS has some significantly more demanding things than plain old transparency going on.
