The survival of coesite in ultrahigh-pressure (UHP) rocks has important implications for the exhumation of subducted crustal rocks. We have conducted experiments to study the mechanism and rate of the coesite→quartz transformation using polycrystalline coesite aggregates, fabricated by devitrifying silica glass cylinders containing 2850 H/10⁶ Si at 1000°C and 3.6 GPa for 24 h. Conditions were adjusted following synthesis to transform the samples at 700–1000°C at pressures 190–410 MPa below the quartz–coesite equilibrium boundary. Reaction proceeds via grain-boundary nucleation and interface-controlled growth, with characteristic reaction textures remarkably similar to those seen in natural UHP rocks. We infer that the experimental reaction mechanism is identical to that in nature, a prerequisite for reliable extrapolation of the rate data. Growth rates obtained by direct measurement differ by up to two orders of magnitude from those estimated by fitting a rate equation to the transformation–time data. Fitting the rates to Turnbull's equation for growth therefore yields two distinct sets of parameters with similar activation energies (242 or 269 kJ/mol) but significantly different pre-exponential constants. Extrapolation based on either set of growth rates suggests that coesite should not be preserved on geologic time scales if it reaches the quartz stability field at temperatures above 375–400°C. The survival of coesite has previously been linked to its inclusion in strong phases, such as garnet, that can sustain a high internal pressure during decompression. Other factors that may play a crucial role in preservation are low fluid availability — possibly even less than that of our nominally “dry” experiments — and the development of transformation stress, which inhibits nucleation and growth. These issues are discussed in the context of our experiments as well as recent observations from natural rocks.