Most physiologists and cell biologists around the world agree that homeostasis is a fundamental tenet of their disciplines. Nevertheless, a precise definition of homeostasis is hard to come by. Often times, homeostasis is simply defined as "you know it when you see it". Mathematical treatments of homeostasis involve studying equilibria of dynamical systems that are relatively invariant with respect to parameters. However, physiological processes are rarely static and often involve dynamic processes such as oscillations. In such dynamic environments, quantities such as average values may be relatively invariant with respect to parameters. This has recently been called as `homeodynamics'. In this talk, we will present a general biomolecular feedback system involving two time scales that elicits homeostasis in the average value of a relaxation oscillator. The key point is that homeostasis manifests when measuring the slow variable and is not apparent in the fast variable. We demonstrate this in the Fitzhugh-Nagumo model and then describe the pheomenon in the Chay-Keizer model-- a model for describing bursting in electrical activity and calcium oscillations in pancreatic beta cells. We briefly discuss a generalization to a system with multiple scales.