Naming, predecessor corrections, and debugging

All files generated here use the En01-C name. The author-supplied En01-A file retains older Em02-A names internally; that documentary inconsistency should be corrected in that file without changing its mathematics. This manuscript does not claim to have rewritten the supplied predecessor bytes.

Two corrections to the supplied En01-B source are relevant:

1. For its three separations and weights, the moments are

   $$\sum _i{w}_i(d_i/R{)}^3=\frac{11}{32},\sum _i{w}_i(d_i/R{)}^5=\frac{239}{512}.$$

   The printed values $15/32$ and $435/1024$ in its rational contrast proof are incorrect. With the corrected moments,

   $$-\frac{3}{16}\frac{11}{32}+\frac{1}{160}\frac{239}{512}=-\frac{5041}{81920},$$

   so the stated final sphere contrast is unchanged.

2. The finite-sample log-inversion budget requires an exact response law $q=a+\kappa V$, or an explicit additional bound on response misspecification. A deterministic physical attenuation residual does not automatically absorb a detector-response error before the logarithm. Here the positive certificate uses the finite sensor record directly and bounded complete-law errors, avoiding that identification.

The following distinctions are enforced throughout:

1. Point-pair coherence, projected packet coherence, and measured fringe visibility are not identified.

2. One common physical apparatus is used under both hypotheses.

3. Outcome windows and electronic rules are fixed before a run.

4. Accepted states are not substituted for unnormalized instrument blocks in attempted-run likelihoods.

5. White and colored support noise use different transfer formulas.

6. Nonmatching generators are not automatically distinguishable after a restricted detector.

7. The full regularized model need not preserve Gaussianity even though its second-moment equations match the quadratic comparison.

8. Finite sensing includes measurement backaction.

9. The four-time variation is a declared design extension, not a reproduction of the reference's one timing choice.

10. Mean cancellation does not imply a bounded-statistic gap without uniform tails.

11. Cooling failures and state-dependent missing records are not assigned an independent-erasure factor without justification.

12. All sample counts refer to attempted runs, not accepted fringes.