The paucity of high-silica rhyolite in volcanic arcs and its restricted occurrence as scattered aplite dikes throughout arc granitoid batholiths suggests there is a mechanism that prevents high-silica (near-eutectic) rhyolite melts from coalescing and erupting as discrete liquids at subduction zones. In contrast, large volumes (≥100 km3) of high-silica rhyolite, including those that are relatively cold (700–750 °C) and hydrous, erupt in extensional tectonic settings. The emerging question is: What controls the eruption of high-silica rhyolite? Through available phase-equilibrium and kinetic studies from the literature, it is shown how high-silica (near-eutectic) melts that segregate when saturated with an H2O-rich fluid will undergo degassing-induced crystallization immediately upon ascent and reach their solidus before eruption. In contrast, high-silica (near-eutectic) melts that segregate in the absence of a fluid phase will ascend into a superliquidus condition, because the liquidus temperature drops with decreasing pressure under fluid-absent conditions. Continued ascent leads to fluid saturation, causing melts to inevitably cross their liquidus from a superliquidus condition. A kinetic delay to nucleation and crystal growth enables these rhyolites to transit through their narrow liquidus-solidus interval without fully crystallizing, enabling eruption. The hydrous nature of arc magmas at the time of emplacement into the upper crust, leading to late-stage rhyolitic interstitial melts that are saturated in an H2O-rich fluid, may be the fundamental reason why high-SiO2 interstitial melts rarely coalesce into larger magmatic bodies at higher levels in the crust after segregation, and why the average upper continental crust has a bulk composition of dacite and not rhyolite.