GCFT CDN-168: The Speed of Coherence — v* Derived, Canonical Kinetics Confronted

CDN-168 pays theory debt D5 and opens a fork at the foundation of the chain. Time-domain sweeps of the CDN-166 drag peak across source sizes σω ∈ [1, 32] find the window scale quasi-universal at v* = (0.15±0.03)·c_Ξ, refuting the 1/(σω) source-size hypothesis: v* is a property of the coherence medium (the relativistic turnover of the gapless branch), not of the system. Combined with the CDN-167 Local Group interval v* ∈ [98, 174] km/s, this requires a coherence propagation speed c_Ξ ≈ 650–1450 km/s — an effective sound speed c_s ~ (2.2–4.8)×10⁻³ — while the canonical Lorentz-invariant kinetic term assumed since the founding of GCFT yields v* ≈ 45,000 km/s, missing by a factor ~300. The pre-registered fork: Branch A (window real) — the Ξ kinetic sector is non-canonical and, since the same field is the dark energy, predicts clustering dark energy with c_s(DE) ~ (2–5)×10⁻³, a cross-sector falsifier requiring Local Group kinematics and DE perturbation constraints to return the same number (currently allowed); Branch B (canonical kinetics) — the window sits at 0.15c, is irrelevant at galactic velocities, and the CDN-161 stream tension returns in full. Ledger: D5 paid (v* derived); D6 opened (compute the spintessence background sound speed and check against the interval). Sweep script, results JSON, and figure included. By Nicky Hacquier (Aporion Dynamics) — Project Aporion; follows CDN-167 (DOI 10.5281/zenodo.20636296).