Galaxy clusters point. Binggeli noticed in 1982 that cluster major axes tend to align with their neighbors, and modern measurements have stretched the effect across distances that break the standard toolkit: cluster orientations correlated over 200 to 300 comoving megaparsecs, brightest cluster galaxies aligned with their host cluster axes at one-in-a-million significance, and the whole architecture already fully in place at redshifts above 1.3 (Binggeli 1982; West et al. 2017).
Each element defeats a different part of the standard account. ΛCDM simulations produce orientation coherence only to roughly 30 megaparsecs, the reach of tidal coupling along filaments; the observed scale exceeds it by a factor of ten. Tidal torquing needs time to reorient structures, with dynamical friction timescales of gigayears; alignment complete at z above 1.3, only four gigayears after the Big Bang, leaves no schedule for gradual assembly. And the BCG-cluster alignment ties the largest scales to the smallest: the central galaxy knows its cluster's orientation, the cluster knows its supercluster's, across a chain of scales no local mechanism connects. The model can produce weak local alignment; it cannot produce deep, early, scale-spanning orientation memory.
The standing strengthens with redshift coverage: each deeper cluster sample finds the alignments earlier and stronger, the opposite of what gradual tidal assembly predicts and exactly what an initial condition would look like.