Averaged statistics certify homogeneity near 100 Mpc/h, yet the universe keeps producing individual coherent structures far above it: the Sloan Great Wall (400 Mpc), the Huge-LQG (500 Mpc/h), the Giant Arc (1 Gpc), the Hercules-Corona Borealis Wall (up to 3 Gpc). The ΛCDM consistency ceiling for real structures is about 370 Mpc (Yadav et al.), exceeded severalfold by a growing inventory. The averages and the objects disagree about the same sky.
The model generates structure from a Gaussian random field whose coherence dies at the homogeneity scale, so statistical convergence and a hard ceiling on individual objects are the same prediction. It has no mechanism that can leave the averages untouched while building rare gigaparsec features, forcing it to litigate each discovered structure as artifact or fluke.
SCT replaces the hot-dense-center with a superluminal collision cascade, and the two clashing observations become two different layers of one process. Statistical homogeneity above about 300 Mpc is a derived property: thermalization treats the overlap volume as a unit, so averaged measures converge just as surveys find (P6). But the deposition itself had geometry: the first and largest collision stages laid down coherent features at the scale of the colliding pockets, with characteristic size Λ_max of about 2 R_pocket, near 5 Gpc, in ring-and-filament morphology from shock-compressed shells (P55, P33, P34).
So the averages converge and the giants exist, without contradiction: homogeneity is a statement about the thermalized mean, the giant structures are the rare, countable scars of the first deposition stages, populating exactly the band between the Yadav ceiling and Λ_max where ΛCDM forbids structure and SCT requires a few. The model discriminates by inventory shape: SCT expects a sparse population of coherent features up to several Gpc with ring and arc morphology, not a continuum of excess clustering that would drag the averaged statistics off their convergence.
This is the same first-stage geometry that produces the Giant Arc and Big Ring individually and the low-multipole anomaly family in the CMB. There is no need to relitigate every structure as a percolation artifact, and no cosmological principle is harmed: it governs the infinite-space average, not our patch's particulars.
The registered kill: DESI, Euclid, and Roman finding no power excess at wavenumbers below 0.01 Mpc⁻¹ beyond cosmic variance, and dissolving the confirmed giant structures into selection or projection artifacts, would remove the first-stage deposit signature. Conversely, structures discovered above the Λ_max scale of about 5 Gpc, or a continuum of large-scale clustering that breaks the averaged homogeneity convergence, would violate the SCT deposit spectrum from the other side.