33.66 turns two-station common-term readout into a retainable propagation audit: if clustered same-window events above Pth produce the same nondispersive step at two stations on the same axis, the inter-station delay ΔT must scale linearly with extra path length D, define a stable v_c, remain approximately unchanged across E_miss or |p_miss| bins, weaken off-axis or below threshold, and keep beta-minus/beta-plus differences confined to step polarity rather than to the distance law itself.
Use this section as a compact machine-readable EFT reference.
33.66 turns the nondispersive common-term step into a two-station propagation court. The admissible claim is not that any synchronous readout means transport, but that above-threshold same-window events show the same step at two stations on the same axis with one reproducible delay structure rather than a local crosstalk pattern.
Measurement preserves one two-station transport ledger. The hard outputs are Δt_common,1 and Δt_common,2, onset times t1 and t2, step heights H1 and H2, the inter-station delay ΔT=t2−t1, and a multi-configuration scaling set in which D=d2−d1 is varied to estimate a stable propagation parameter v_c. Delay curves binned by E_miss or |p_miss| test energy independence, while only clustered same-window events in the P ≥ Pth regime and inside a frozen axis cone θ≤θmax may enter the court.
The workflow freezes geometry, timing, and extraction before any verdict. Two stations are placed on the same axis line at preregistered d1 and d2, one shared time reference or calibrated synchronization method ties them together, both stations use the same bandwidth, sign convention, filtering, and fit aperture, the t1/t2 extraction algorithm is frozen before data taking, at least three D configurations are acquired with P scans, and distance, channel, and P-level labels stay blinded during delay extraction.
Controls must destroy false propagation structure rather than merely soften it. Under P=0 there may not be stable same-window steps with a reproducible distance-linear delay; below threshold the two-station coincidence rate and the ability to fit ΔT versus D must drop sharply; off-axis placement or off-axis event selection must weaken the court; station-label or event-window permutations must collapse the scaling; and the common-term step must remain nondispersive across bandwidths.
Support requires same-window steps at both stations with a stable ΔT peak, a linear ΔT–D law across at least three configurations, a stable v_c across batches and days, and approximate invariance of ΔT across E_miss or |p_miss| bins. Falsification follows from no reproducible coincidence, no linear distance law, energy dependence or dispersion-like behavior, or null controls that reproduce the same-strength scaling. The named adversaries are clock synchronization drift, electromagnetic crosstalk or path leakage, and geometric alignment or distance-calibration error.
So the chapter closes one retained two-station common-term propagation-scaling ledger aligned with the V03/V08/V09 audit lane. v_c may survive only as a geometry- and chain-specific propagation parameter for this object court, not as a new universal speed, medium constant, or canon-core transport ontology. Its onward value is to hand a spatial footprint card to 33.67 and a step-sequence grammar to 33.68.