Pre-caldera high-silica rhyolites of Glass Mountain, California erupted episodically from 2.1 Ma until the catastrophic eruption of the Bishop Tuff at 0.74 Ma. The lavas are extremely evolved, with Rb/Sr ratios between 128 to 3640, the latter being the highest recorded from a volcanic rock. Glass separates from pre-1.2 Ma lavas define two geographically controlled RbSr isochrons. Lavas adjacent to the current caldera rim define an isochron age of 2.047 ± 0.013 Ma with an initial ratio of 0.7063 ± 2, and lavas more distant from the caldera define an isochron of 1.894 ± 0.013 Ma with the same initial ratio. The isochrons are consistent with the magmas forming within 26 ka, which implies a minimum magma production rate of 0.75 × 10-3 km3/yr over this period. New 40Ar39Ar ages on sanidine and biotite have established that lavas defining each isochron were erupted over a long time interval, the isochron ages being up to 360 ka older than the youngest eruption age. RbSr isotope data are reported for minerals from three lavas with eruption ages of 1.990 ± 0.012, 1.866 ± 0.014 and 1.686 ± 0.011 Ma. Petrographically early apatite inclusions in biotite and biotite inclusions in feldspar and quartz have glass-mineral RbSr ages that are indistinguishable from the relevant regional isochron. Sr diffusion in feldspar is slow at the magmatic temperatures inferred for Glass Mountain rhyolites (∼ 700°C) such that over 0.5 Ma the cores of large feldspars ( > 1 mm) will retain > 99.9% of their original Sr. The cores of sanidine and plagioclase yield glass-mineral ages that are up to 300 ka older than eruption ages. Feldspar rim ages for two samples are indistinguishable from eruption ages. The rims of sanidines and plagioclases from the third sample are 110 and 280 ka older than the eruption age and 180 and 20 ka younger than the cores. These mineral age data probably reflect the combination of extended periods of mineral growth and partial isotopic exchange with the host liquid during protracted residence in a magma reservoir. However, the Ar and Sr isotopic data for biotite phenocrysts are consistent with the presence of a significant component that is recycled from earlier magmatic pulses. Due to the extreme Rb/Sr ratios of the rocks and minerals it is possible to very precisely resolve the time difference between the formation of different phases, assuming that they crystallised from the same host magmas (e.g., Tplagioclase - TFeTi oxide = 6.8 ± 0.1 ka) and the maximum time taken to form a phase (e.g., Tplagioclase core - Tplagioclase rim = 32.3 ± 0.2 ka). The timescale for mineral growth is shorter in the chemically more evolved and crystal-poor lavas, consistent with these magmas having resided at higher levels in the magma chamber with shorter residence times and being more liable to extrusion. By using mineral inclusion relationships, average mineral growth rates are estimated to be between 7 × 10-13 and 8 × 10-14 cm/s. These values are significantly lower than those measured in basaltic systems and probably reflect a combination of the slow cooling rate of the Glass Mountain magma chamber(s) and the highly polymerised nature of high-silica magmas.