p22: evanescent components (EV-M2), evanescent counterterms (EV-CT) and the runtime d guard

revision (disclosed). p22 carries one disclosed contract revision, listed in p22_conventions.CONTRACT_REVISION: (i) evanescent counterterms are admitted outside the p14r 4D domain, and (ii) code/p5_entry/dred.py gains one delimited additive block (the runtime $d$ guard). Outside the block the file is byte-identical to the p18 release; contract_registry.dred_revision checks this, and the p22 gate pins the exact sha256 of the revised file.

-M2: the matching statement. XVI-O has no evanescent counterpart, so the eight evanescent ($\tilde{g}$) structures of D11 cannot be matched to XVI-O. We tested three candidate statements exactly, with $d$ opaque: (a) {match a $d$-dimensional continuation of the XVI-O kernel} is refuted for both natural continuations ($g_4\to \hat{g}$ and $g_4$ kept), because the exact D11 tensor differs from either continuation of its own 4D part on $\stackrel{~}{}$ external legs. XVI-O defines no other continuation, so (a) is not a usable statement. (b) Pure local evanescent operators: proven. Every evanescent pole (D01, D03, D11) is a polynomial in $p$ and independent of $d$ after pole extraction. (c) Vanishing in the renormalised result: proven in a stronger form. After subtraction, every evanescent component is identically zero on 4D (hatted) external legs at every $d$, not merely at $O(ϵ)$. The scope is D11_matching_verified_scope = “4D components: M2 discharged (P18_D11_M2, unchanged); evanescent components: EV-M2 proven; NOT matched to XVI-O”.