NGC 1052-DF2 and DF4 show globular-cluster dispersions of 8 to 10 km/s, fully explained by their stars alone: no dark matter. ΛCDM must form these galaxies inside halos and then remove the halos by tidal stripping tuned to take all the dark matter while sparing the diffuse stars and their oddly luminous clusters, twice in one group. MOND overshoots in the other direction and needs an external-field rescue.
If dark matter is a particle substance bound in halos that host galaxy formation, a galaxy without it is an exception requiring surgical history: per-object stripping narratives with per-object luck. The model has no parameter that makes dark matter signatures conditional; the halo either survived or the galaxy should not exist.
SCT replaces the dark matter particle with coherence: apparent dark matter is the constructive superposition of comoving sources, governed by the coherence function C = exp(-σ_v²/v_cross²), which interpolates continuously from full amplification A near A* = 5.970 in coherent systems down to A = 1 when coherence is destroyed (P50, P52, P53). Dark matter deficiency is then not a missing substance but a destroyed phase relationship. Tidal interaction with NGC 1052 does the one thing the mechanism cares about: it stirs the internal velocity dispersion relative to the shared bulk motion, driving σ_v/v_bulk up and C toward zero. The galaxy keeps its stars and clusters and loses only its amplification, exactly the observed configuration, with no fine-tuned selective stripping required.
The rule and the exceptions come from one formula. Ordinary galaxies are coherent ensembles and show A above 1; tidally processed ultra-diffuse galaxies are decohered and show A of 1; and the prediction in between is continuous: A_obs should decrease monotonically with σ_v/v_bulk across the UDG population. Two such galaxies in one group is not compounded luck but shared environment, since one tidal field decoheres its neighborhood. MOND's overshoot is also diagnostic: an acceleration-scale law cannot turn off, while coherence can.
This is the same C-function physics that sets cluster amplification and the radial acceleration relation. There is no need to invoke bullet-dwarf ballistics or distance revisionism.
The registered kill: across twenty or more ultra-diffuse galaxies with measured kinematics (JWST IFU programs in Coma), A_obs must correlate inversely and monotonically with σ_v/v_bulk; no correlation at more than 2σ falsifies the decoherence account of dark matter deficiency. A robustly isolated, dynamically cold UDG with no tidal history and still no dark matter signature would likewise break the mechanism, which requires an agent of decoherence.