To extract the pristine Cosmic Microwave Background, researchers must subtract foreground emission from point sources such as radio galaxies, quasars, and dusty star-forming galaxies. Standard analyses rely on models of source counts and spectral energy distributions derived from Lambda-CDM structure formation simulations and extrapolations from other frequencies. However, there are persistent indications that these models may be overestimating the point source contribution, particularly at high frequencies or small angular scales, leading to "over-cleaning" that suppresses real signal or introduces spurious correlations in the residual maps (Planck Collaboration 2020; Ade et al. 2011).