Debugging summary

The revision makes the following corrections explicit:

1. Class transfer is conditional. A changed source is not proof of a changed free-shell class. The exact same-class transfer is Proposition prop:transfer.

2. The documentary finding is bounded. The supplied retrieval report is not an exhaustive nonexistence theorem.

3. The sharp comparison has a domain. $H{J}_3$ is used only for smooth evaluated sources. No discontinuous multiplier is applied automatically to an unidentified distributional kernel.

4. Temporal support is not compact support. A closed ramp layer is spatially unbounded.

5. Late source agreement is not response agreement. Lemma lem:memory gives an explicit counterexample to that inference.

6. The boundary conclusion is narrow. No boundary action follows from the cited XI-A passage. Weighted representative changes still have (eq:boundaryrepresentative).

7. Frequency factors are separated. $\mathsf{q}$ and ${\mathfrak{d}}_1$ are independent of $\lambda$; $\mathfrak{d}=\lambda {\mathfrak{d}}_1$.

8. The ODE converse has its hypotheses. The switch is real and $\lambda >0$.

9. Polarization is not rejected on typographical grounds. A valid diagonal root identity on a real vector space polarizes by Lemma lem:root-polarization. That scalar operation does not independently verify an unconstructed coordinate-field representation.

10. Later progress is not erased. XVI-D's response and XVI-E's root pairing are recorded. XVI-F-v2 supplies direct auxiliary decay, not a transfer from the root.

11. The integrated support error is corrected. A compact source can generate an affine past tail. Effective switched amplitudes can be compact even when the integrated kernel is not.

12. No stale hypothesis is cited. No null-survival assumption is imposed. The supplied revised XVI-A contains no adopted hyp:null-survival gate.

13. No compilation result is fabricated. This is a single-file pdfLaTeX source. Compiler execution and full-corpus verification are not claimed.