Global climate models generally underestimate the magnitude of precipitation extremes as compared to observations. An approach that has made headway on bridging this gap is superparameterization, wherein the usual statistical representation of the effects of clouds is replaced by actually nesting high-resolution models within the parent climate model. Despite this effort, the microphysical processes occurring within clouds must still be approximated in a statistical manner similar to that mentioned above. Because there are several plausible ways to represent such processes, it is possible that the gap between modeled precipitation extremes and observed ones could be narrowed further, if a certain method is chosen. Here, we examine four microphysical schemes and find, using extreme value theory, that the modeled precipitation extremes can show statistically significant differences, depending on the scheme. To determine whether the effect is direct and local or indirect and dependent on first altering the surrounding large-scale weather systems, a series of shorter simulations were done, during which it was found that the aforementioned statistically significant differences disappear. This suggests that the effects are not direct/local, but rather dependent on indirect feedbacks requiring longer time scales to manifest in precipitation extremes. Despite this conclusion, none of the microphysical representations tested significantly narrowed the gap between the magnitude of modeled extremes and that of ones observed in the contiguous United States (CONUS). Finally, the magnitude and the spatial dependence of the increase in CONUS precipitation extremes from present-day to future, warmer climates are insensitive to the microphysics scheme.