> Making sure they're aren't a bunch of alloc's or GC spent in the hot path of a high performance library is hardly a premature optimization.

You often don't know the what the hot paths are.