[G2] Target-clean propagation discriminator and physical-bridge preflight

[muellerberndt]

Restore an active producer lane for the immutable FZ-11 and FZ-12 propagation discriminators. G1 (#695) remains custody and comparison governance only. This issue may construct source bridges, lower bounds, power, and exposure contracts, but may not rewrite frozen payloads or unseal a comparison.

Exact target-clean mathematics attained:

• the positive-weight carrier-class theorem fixes the negative quartic sign, B0/C4^2 >= 10/21, the rank-six band -16/135 <= B6/B0 <= 16/75, and the common-radius identity 5 D6 B0 = 12 B6 D0;
• the vertex and edge frozen rays are exact members of that class;
• for every finite strictly positive Euclidean tight frame, the normalized complete cosine symbol is the squared norm of an exact sine-feature map;
• that feature map is a contraction, so its auxiliary frequency is globally 1-Lipschitz in the selected Euclidean carrier chart for all momenta;
• exact FZ-11 and FZ-12 bindings give t=4 with prefactor 1/(2a^2) for vertex12 and t=10 with prefactor 1/(5a^2) for edge30;
• Lean, a deterministic receipt, an independently implemented verifier, byte-exact replay, and mutation controls cover the contraction packet.

Scope:
The unit constant is a certified upper bound, not a formal optimality result. The receipt contains no public comparison payload. It supplies no physical position or frequency identification, clock, photon or field sector, wave packet or signal front, carrier frame or boost, finite physical scale or positive lower bound, source lag, detector readout, branch exclusivity, or comparison permission. It is not a speed-of-light prediction.

Priority continuation:

1. construct a same-action physical position/frequency/clock attachment;
2. search for a source-derived positive carrier-scale lower bound;
3. freeze source-lag, frame, exact-remainder, nuisance, covariance, and decision rules before any new exposure;
4. derive blinded power and coverage for the complete linked coefficient manifold;
5. retain UHE threshold work as secondary until its lepton and shower response models exist;
6. keep conditional branch failure separate from an OPH-wide verdict unless forcedness and exclusivity are proved.

Required controls: immutable FZ-11/FZ-12 hashes; exposure ancestry; synthetic blinded power; wrong-frame, wrong-normalization, incomplete-covariance, zero-scale, and source-lag controls; independent verification.

Comparison and verdict authority stays with G1 (#695).

Dependency semantics: the target-clean mathematical preflight may complete without E2, E5, or F1. A physical propagation promotion additionally consumes the source-realized net, clock, and spacetime receipts supplied by those lanes.

Depends on: none.
Wave: standing.

[muellerberndt]

Final audit hardening landed in d573a011. The carrier theorem is restricted to the declared positive-weight scalar cosine class whose full spatial symbol is the orbit sum. The issue text distinguishes the exact analytic/Python certificate from the Lean rational implication skeleton, and records the numerically independent raw-cosine test route. The postdiction ledger and public summaries were regenerated. Frozen FZ-11/FZ-12 bytes are unchanged; source selection, physical field and clock attachment, finite scale, readout, exclusivity, and powered comparison remain the work of this open lane.

[muellerberndt]

Audited implementation checkpoint: f55d5bf on main.

The new exact packet proves that every normalized complete positive tight-frame cosine symbol is the squared norm of its sine-feature map and that the resulting auxiliary frequency is globally 1-Lipschitz in the selected Euclidean carrier chart. FZ-11 and FZ-12 have exact vertex12/edge30 bindings, a deterministic receipt, an independent implementation, byte-pinning, and mutation controls.

The result is an upper bound on an auxiliary carrier coordinate. It is not a physical signal-speed result or a frozen discriminator. The position/frequency/clock, field, signal-front, frame, scale/lower-bound, detector, exclusivity, and comparison bridges remain open, so #704 remains open.