[Koying]

Thought it was generally accepted among people understanding a GL pipeline that forcing a CPU to GPU sync is bad for performance as GL is async most of the time.

I don't get that point. usleep does not halt the GPU, and there is a GPU to CPU sync point when doing the GL buffer swap
In neither approach is the GPU waiting, but the CPU is with both, just that the usleep approach is platform-independent.

EDIT: Well, the GPU *is* waiting, but usleep is making it explicit, while SwapBuffer+vsync makes it implicit.
Result is the same, just that the multiple renderers *not* having implicit each have to reimplement it