High-silica rhyolite, which has a narrow liquidus-solidus interval, is the most differentiated silicate magma type on Earth and makes up some of the largest explosive eruptions, including those at Yellowstone (WY) and Long Valley (CA). An outstanding question is why high-SiO₂ melts are so voluminous in extensional tectonic settings and so scarce as an erupted whole-rock magma type at subduction zones. Although aplite dikes of high-SiO₂ rhyolite composition are ubiquitous in granitoid batholiths (dacite-andesite composition), they rarely erupt and instead invariably freeze during ascent. Their failure to erupt is understood from the work of Tuttle and Bowen (1958): when H₂O-rich fluid-saturated melts decompress, they degas dissolved H₂O, which drives their liquidus to higher temperatures and promotes crystallization. The next question is why similarly hydrous high-SiO₂ rhyolites (e.g., those from Long Valley, CA) successfully erupt. One hypothesis is that hydrous, high-SiO₂ rhyolite must segregate from their crystalline mush source under fluid-undersaturated conditions. If so, then during initial ascent along a fracture, decompression will drive liquidus temperatures lower, leading to super-liquidus conditions in the melt. Continued ascent eventually leads to fluid saturation and H₂O degassing, causing the high-SiO₂ rhyolite to eventually cross its liquidus, but from a super-liquidus condition (i.e., free of nuclei/crystals). A kinetic delay in nucleation is thus expected, which allows an effective undercooling (∆T_eff = T_melt-T_liquidus) to develop. Experiments from the literature show that ∆T_eff values of ~20-100 degrees, difficult to attain in a stalled magma reservoir, lead to low nucleation rates and high crystal growth rates, and thus the formation of large, sparse phenocrysts with diffusion-limited growth textures. This is precisely what is found in obsidian and pyroclastic deposits from both Long Valley (CA) and Yellowstone (WY). An examination of phenocryst textures, abundances, and compositions from obsidians and pumice clasts from both Long Valley and Yellowstone are shown to be consistent with crystal growth during ascent (and not in a stalled magma reservoir) and that melt segregation from their respective source regions must have occurred under fluid-undersaturated conditions.