33.68 turns amplitude scaling into a translation court: with clustered same-window events above threshold, one frozen chain, one frozen near-axis and integration aperture, and at least three blinded distance configurations, |H₀(d)| must fall approximately as 1/d, |H₀(d)|×d and the transverse squared-profile integral F must remain stable within uncertainty, the pattern must collapse under pulse-off / below-threshold / off-axis / label-permutation controls, and cross-band readout may not introduce a dispersion-like rescaling; under V03/V08/V09-compatible translation, this remains a geometric amplitude and integration ledger rather than a conserved-medium verdict.
Use this section as a compact machine-readable EFT reference.
33.68 raises the common-term step from timing and footprint into an amplitude court. The admissible claim is narrower than a transport ontology: above-threshold same-window events must show a geometric amplitude law across distance rather than a local chain artifact or trigger-synchronized common mode.
The hard ledger couples three objects under frozen apertures: the near-axis step height H₀(d) within ρmax, the transverse profile H(ρ, d) with θ₀ or θ₅₀ support fits, and the flux-like quantity F obtained by integrating the squared profile under one fixed ρint. Binning in pmiss∥ or Emiss may change only the normalization, not the distance-scaling exponent, and cross-band readout may not re-scale the law or flip its sign.
The workflow freezes geometry before verdict. At least three distance configurations are deployed along one consistent axis, the same chain, sign convention, filtering rule, and step extractor are reused, transverse array data are taken in the same runs so H(ρ, d) and F are computed together, and distance, β-channel, and power-bin labels remain blinded until only the preregistered inverse-distance and conservation tests are performed.
Controls must break fake scaling instead of merely inflating uncertainty. Under P=0 or P<Pth the inverse-distance relation and F stability should weaken, off-axis relocation beyond the Chapter 67 cone limit should sharply reduce the on-axis law, distance-label or event-window permutation should destroy the apparent constancy of |H₀(d)|×d, and changing bandwidth or band may not induce a new exponent or a dispersion-like sign flip.
Support requires inverse-distance decay of |H₀(d)| across at least three distance settings, agreement of |H₀(d)|×d within uncertainty, cross-distance stability of F compatible with the Chapter 67 profile collapse, and null controls that do not reproduce the same-strength pattern while nondispersion remains intact. Falsification follows from no stable 1/d law, no F conservation, comparable null effects, or bandwidth-dependent re-scaling. The named adversaries are station sensitivity and noise-floor mismatch, axis-definition drift, and integration-aperture or window-function bias.
So the chapter closes only one translated geometric amplitude and flux-like readout ledger aligned with the V03/V08/V09 audit lane. H₀·d constancy and F stability may survive only as geometry and integration bookkeeping, not as a conserved-medium ontology or a canon-core propagation verdict. Its clean onward value is to hand a hardened propagation-scaling grammar to the one-map multi-probe court of 33.69.