Bin size vs. rms plots for a subset of the best-fitting models from our grid. This displays one line for every fit that used an exponential trend and quadratic limb darkening, then either an Matérn 3/2 kernel (dots) or no overlying GP (crosses). The 1σ, 2σ,  and 3σ contours for expected ﹩\sqrt{N}﹩ binning are shown via gray contours. We see that the residuals of the GP fits consistently bin down faster than white noise after the bin size exceeds roughly 10 minutes, but that the no-GP fits immediately diverge from white noise behavior. We also note that the cadence the fit was performed on does not seem to influence residual binning behavior.