Keck and VLT archival quasar spectra show the fine-structure constant varying as a spatial dipole at 4.2σ across 300 absorbers (Webb et al. 2011), with the axis lying near other anomaly directions, while supercalibration studies expose wavelength distortions of the dangerous form and modern ESPRESSO precision finds no variation in the systems it has covered. A constant of nature stands accused of having a direction, on contested evidence.
The standard framework requires alpha to be one number everywhere: its constancy is an axiom inherited from a homogeneous, structureless vacuum. The model therefore offers no scale for any variation, no preferred axis if variation exists, and no reason the claimed dipole should align with anything; it can only hope the systematics win.
SCT gives alpha permission to vary and tells it how: the framework's registered prediction is spatial variation proportional to the local coherence field, delta-alpha/alpha tracking delta-C/C, because the electromagnetic sector's effective couplings ride the same carrier-interference structure that sets gravitational coherence (the Paper 17 ledger entry, grounded in the superposition premises P50, P53). The variation is therefore structured, not random: largest between regions of contrasting coherence, organized along the gradients of cosmic structure, and naturally dipolar at the survey scale, with a predicted axis tied to the large-scale coherence gradient, the same family of directions as the bulk-flow and dipole anomalies the claimed alpha axis already neighbors.
The framework's honest layer: the predicted amplitude is environment-differential, so modern null results on individual sightlines do not yet contradict it, while the archival dipole, if calibration-driven, would simply leave the prediction open rather than confirmed. The decisive configuration is directional coverage at ESPRESSO precision: variation correlated with traversed coherence contrast, not with telescope or epoch.
This is the same coherence field that produces the dark-matter phenomenology and the G_N covariation prediction (delta-G/G = 2 delta-alpha/alpha, exactly). There is no need to defend absolute constancy as an axiom; SCT replaces the axiom with a measurable law.
The registered kill: alpha confirmed constant across environments of contrasting coherence at 10⁻⁶ precision, with systematic ESPRESSO and ELT coverage of the claimed axis finding no structured variation, falsifies the coherence-coupling prediction. The companion ratio is the sharper blade: if variation is detected, delta-G_N/G_N must equal exactly twice delta-alpha/alpha; any other measured ratio breaks the carrier-structure derivation outright.