Modern large-volume cosmological simulations rely on stellar
feedback-driven outflows to solve the 'overcooling' problem, but they must
incorporate such outflows via highly uncertain sub-resolution
prescriptions. In contrast, state-of-the-art simulations with resolved
stellar feedback generate outflows self-consistently. I will discuss the
underlying physics of such outflows and describe how simulations and
theoretical models for outflows can be tested using future radio/(sub)mm
facilities.