Planck disagrees with itself: ΛCDM parameters fit to multipoles below about 800 differ at 2 to 3σ from those fit above it, with omega_c split at 2.5σ and H₀ pulled several km/s/Mpc apart (Addison et al. 2016). The high-multipole half prefers extra lensing smoothing, the A_lens excess, and absorbing it drags the high-ell parameters away from the low-ell ones. Same sky, same instrument, two universes.
The model assumes one parameter set governs every angular scale and that all peak smoothing comes from lensing by catalogued matter. An unmodeled smoothing contribution that lives where lensing lives, in the damping tail, then contaminates only the high-multipole fit, and the model has nowhere to put it except into scale-dependent parameter shifts it must call coincidence.
SCT replaces the hot-dense-center with a superluminal collision, and the matter potential with Φ_eff(r) = Φ_local(r) + Φ_mesh(r) (P51): a coherent parent-frame mesh that adds real lensing convergence along every sightline (P50, P52, P53). Mesh convergence smooths the acoustic peaks exactly as lensing does, so it enters the likelihood only through the lensing-sensitive high multipoles. A mesh-blind fit to that range must misassign the excess to the parameters that control smoothing, pushing omega_c up and H₀ down, while the low-multipole fit, where smoothing barely registers, stays clean. The internal split is the mesh sorting itself by angular scale.
The diagnostic ΛCDM cannot explain is the one SCT predicts: a single phenomenological smoothing amplitude reconciles the halves, because the mesh is a single physical smoothing amplitude. Its predicted size, A_lens of about 1.17 from the superposition boundary condition integrated over the lensing kernel, matches the PR3-era excess that drove the split. The mechanism costs nothing new: GR is exact at N = 1, and the background expansion carries the usual dynamical Λ_eff (P17) as a secondary modulation.
This is the same coherent term that appears as the lensing-amplitude anomaly, the closed-universe pull, the S₈ deficit, and the growth-index excess: five readings of one omission. There is no need to invoke organized statistical flukes or an instrument systematic no one has found.
PR4/NPIPE reprocessing shrinks the smoothing excess to 1 to 2σ and ACT DR6 is internally consistent, so the mesh amplitude this mechanism requires is under live observational pressure; CMB-S4's damping-tail measurement is the decisive test.
Two kills. CMB-S4 converging on A_lens = 1.000 +/- 0.005 removes the mesh smoothing and with it this account of the split. And the split must be lensing-shaped: if the low-versus-high multipole parameter shifts persist at full strength after marginalizing over a free lensing amplitude, the discrepancy is not an absorbed smoothing term and the mesh reading fails.