How exceedingly large, volcanic supereruptions begin provides crucial information on the storage, ascent and release of silica-rich magma in catastrophic events. Initial fall deposits of the 2.08 Ma, 2500 km3 Huckleberry Ridge eruption are multiply bedded and in several places contain reworked intervals, indicating time breaks in the opening phases of the eruption. A 2.5 m section of these fall deposits was sampled at nine levels below the earliest ignimbrite (member A) at Mount Everts (Mammoth, Wyoming). We analyzed major and trace elements and volatiles in quartz-hosted melt inclusions (MIs), reentrants (REs; unsealed melt inclusions) and associated obsidian pyroclasts (thick-walled shards) to establish quartz crystallization and storage depths and melt compositional groupings. Systematic relationships between Rb and other incompatible elements (U, Cl, B) indicate ∼55% fractional crystallization between the least and most evolved glass compositions. In contrast, H2O concentrations in MIs show scattered relationships with trace elements and are interpreted to reflect variable loss of H2O by diffusion through the quartz host during magma ascent. The wide H2O variations (1.0–4.7 wt.%) in MIs from individual fall horizons imply as much as ∼14 days of diffusive loss, reflecting highly variable and surprisingly slow decompression conditions. Water and CO2 gradients in reentrants, however, are consistent with final ascent times of <1 to 4 h (ascent rates of ∼0.3–1.5 m/s), similar to those represented by MIs that we infer to have experienced little to no diffusive H2O loss. The wide range of ascent rates for co-erupted crystals mirrors that of intermittent explosive activity at Mount St. Helens in summer 1980, and implies that the Huckleberry Ridge magma body was not strongly overpressured at eruption onset. Restored entrapment pressures and geochemical data for MIs provide evidence for six distinct populations of quartz that originally crystallized in geochemically distinct magma domains. The compositions of REs and obsidian pyroclasts, by comparison, show that by the onset of eruption, the quartz had been brought together into three discrete magma bodies, which we interpret to have been cupolas on the roof of the main magma body. These cupolas were erupted sequentially and episodically from separate vents to generate the fall deposits before escalating activity led to generation of voluminous pyroclastic flows, and this pattern of activity suggests that tectonic triggering may have destabilized multiple magma bodies. Supereruptions as large as the Huckleberry Ridge event may start hesitatingly if the parental magma bodies are not strongly overpressured, with small-scale episodic activity that is modulated by external controls that may leave no other geological evidence for their presence.