The cosmological principle requires the universe to become statistically homogeneous above some scale, and the textbook answer is that it does, near 70 to 100 Mpc/h, where fractal analyses of galaxy surveys show the correlation dimension converging to 3 (Scrimgeour et al. 2012; eBOSS quasar multifractal analyses). The debate is that the universe keeps building structures larger than the homogeneity scale that certifies it: the Sloan Great Wall at about 400 Mpc, the Huge Large Quasar Group near 500 Mpc/h, the Hercules-Corona Borealis Great Wall estimated at 2 to 3 Gpc, the Giant Arc at 1 Gpc, and the Big Ring at 400 Mpc diameter. Yadav et al. derived a ΛCDM consistency ceiling near 260 to 370 Mpc for genuine structures; multiple confirmed objects exceed it severalfold.
The contradiction is statistical versus particular: averaged two-point measures converge to homogeneity on schedule, while individual coherent structures, found by morphology rather than averaging, keep appearing at scales the model's Gaussian random field should essentially never produce. ΛCDM has no native way to reconcile the two; it either disputes each structure's reality (chance projection, percolation artifacts of friends-of-friends linking, a posteriori statistics) or accepts isolated multi-sigma flukes one at a time, while the inventory of gigaparsec-scale candidates grows. The fractal-dimension defense and the giant-structure offense talk past each other because they measure different things.
The standing is contested at the definitional level: what counts as a structure, how to assess significance post hoc, and whether the cosmological principle is a statement about averages or about the largest allowed coherent objects. Forthcoming DESI and Euclid catalogs, an order of magnitude deeper, will either dissolve the giant structures into selection artifacts or make the inventory undeniable.