Now, I'm making this up as I go, but I think the situation's even worse than that.
Suppose we postulate a universe that has gravity and massive objects (for example, our universe). Now, that means space isn't flat and I think that means that the relationship between radius and circumference for a particular patch of space isn't exactly pi.
Of course, to the extent that you know the mass distribution, then you can compensate for the curvature of space (assuming that you're far, far better at math than I), but I suspect that the uncertainty principle keeps you from ever perfectly knowing the distribution of masses in a volume of space, so you're never perfectly know the curvature, so pi's utility may run out even sooner.
ETA: Or maybe it's a really small effect. I don't know.