Damping-tail cosmology rides on a subtraction: unresolved point-source power, extrapolated from bright counts, fit jointly with the cosmology. The treatments differ between Planck, SPT, and ACT pipelines, shift the inferred parameters when varied, and contribute to the documented inter-experiment offsets, while JWST keeps discovering faint populations more abundant than the extrapolations assume. The model's precision frontier rests on an undercounted ledger.
The faint-end extrapolations inherit the model's assembly history: source counts below threshold follow from the same hierarchical population synthesis under strain everywhere else. If the universe front-loaded more faint emitters than hierarchy builds, every template calibrated on the bright minority systematically misweighs the unresolved sky.
SCT's ledger says the faint end is fuller than hierarchical extrapolation allows: collision seeding front-loads compact accretors and early stellar systems in abundance (P46, P56), the demographic JWST now confirms as the Little Red Dots and the bright-end luminosity-function excess, with the pre-existing matter context deepening every epoch's emitter census (P25). The point-source foreground is therefore predicted to be misestimated in exactly the observed manner: more faint sources than the counts extrapolate, with redder, dustier, higher-redshift spectral mixes than bright-source calibrations project between bands, leaving residual power that pipelines absorb differently and parameters that drift with the treatment.
The framework converts the soft spot into a cross-check: the same seeded census that explains the optical and infrared background excesses and the gamma-ray attribution gap must appear here as foreground residuals correlated between frequency channels in the seeded population's spectral pattern, and the damping-tail parameter offsets between experiments should shrink as deeper source counts replace extrapolation with measurement. The connection to the lensing-smoothing family is direct, since mis-subtracted source power and mesh smoothing both live in the high-multipole range where the model's internal tensions concentrate.
This is the same deep-ledger accounting behind the cosmic optical background and diffuse gamma-ray entries. There is no need to treat the faint sky as a nuisance parameter when it is a prediction.
The ledger is countable: JWST-depth and SKA-depth source counts establishing that the faint-end populations match hierarchical extrapolations, with CMB foreground residuals dissolving under measured counts and the inter-experiment parameter offsets surviving unchanged, would remove the undercount signature and locate the soft spot elsewhere. The seeded census requires the counts to come in high; counts that come in standard falsify this reading of the foreground problem.