Supernova distances provide a direct probe of cosmic acceleration, constraining dark energy. This leverage increases with survey redshift depth at a rate bounded by the systematic uncertainties. We investigate the impact of a wavelength-dependent, global correlation model of systematics in comparison to the standard local-redshift correlation model. This can arise from subclass uncertainties as features in the supernova spectrum redshift out of the observer photometric filters or spectral range. We explore the impact of such a systematic on ground-based supernova surveys such as Dark Energy Survey and LSST, finding distinctive implications. Extending the wavelength sensitivity to 1.05 μm through ``extreme red'' CCDs can improve the dark energy figure of merit by up to a factor 2.