Mid-ocean ridge hydrothermal systems are known to vent fluids with salinities substantially different from seawater as a result of phase separation and segregation of the resulting vapor and brine phases. Time series of vent temperature and salinity (chlorinity) show that some black-smoker vent fields such as the Main Endeavour Field on the Juan de Fuca Ridge have vented fluids with salinities well below seawater for over a decade, which raises important questions concerning the fate of brines in these systems. One widely accepted model is that high-density brines formed by supercritical phase separation sink to the base of hydrothermal systems, leading to the development of a two-layer system in which a recirculating brine layer underlies a single-pass seawater cell. We first present theoretical arguments to constrain the dynamics of such a deep brine layer in a system still undergoing phase separation, and we conclude that if brines are stored in a basal layer, they are unlikely to convect because they will be stably stratified. One consequence of this result is that the brine layer beneath black smoker systems has to be thin (<10 m) to match the high heat fluxes. However, estimates of the rate at which brines are accumulating in the crust below the main field on the Endeavour segment of the Juan de Fuca Ridge suggest that the brine layer is likely at least 100 m thick. To resolve this apparent paradox, we propose an alternative model. We argue that interfacial tensions between fluid and solid phases will likely favor the segregation of vapor into the main fractures and brine into the smaller fissures and backwaters. This allows the vapor to flow efficiently through the system and transport large heat fluxes while most of the porosity in the lower part of the system fills with brines. It is generally believed that the pressure gradients in mid-ocean ridge hydrothermal systems are close to cold hydrostatic. At the high temperatures and pressures characteristic of the deeper parts of these systems, brines with salinities as high as 20 wt % NaCl have densities around 800--900 kg m-3 and will be buoyant in a cold-hydrostatic system. Rather than sinking to the base of the system, it is possible that brines produced by supercritical phase separation rise slowly until they reach a level of neutral buoyancy as they cool or enter high-permeability regions in which the pressure gradients decrease.