Before large volumes of crystal poor rhyolites are mobilized as melt, they are extracted through the reduction of pore space within their corresponding crystal matrix (compaction). Petrological and mechanical models suggest that a significant fraction of this process occurs at intermediate melt fractions (ca. 0.3–0.6). The timescales associated with such extraction processes have important ramifications for volcanic hazards. However, it remains unclear how melt is redistributed at the grain‐scale and whether using continuum scale models for compaction is suitable to estimate extraction timescales at these melt fractions. To explore these issues, we develop and apply a two‐phase continuum model of compaction to two suites of analog phase separation experiments—one conducted at low and the other at high temperatures, T, and pressures, P. We characterize the ability of the crystal matrix to resist porosity change using parameterizations of granular phenomena and find that repacking explains both data sets well. A transition between compaction by repacking to melt‐enhanced grain boundary diffusion‐controlled creep near the maximum packing fraction of the mush may explain the difference in compaction rates inferred from high T + P experiments and measured in previous deformation experiments. When upscaling results to magmatic systems at intermediate melt fractions, repacking may provide an efficient mechanism to redistribute melt. Finally, outside nearly instantaneous force chain disruption events occasionally recorded in the low T + P experiments, melt loss is continuous, and two‐phase dynamics can be solved at the continuum scale with an effective matrix viscosity.