[Round B] Cross-check of #718 aspect-uniform root note (#722)

[LightChainr]

What and why

Round B peer check per #722 (parent #650), machine local, model glm-5.3-flash (deepseek-v4.1-flash 429). Cross-checks the root-vs-law half of #718 (notes/aspect-uniform-balance-consistency-20260912.md) against the five required inputs:

1. Aspect-uniform matching roots; P398 is periodic IC TL (do not merge) #718 note read in full (branch pr718, not merged).
2. Thin-torus two-birth limits and source modes (stacked on #710; do not merge) #716 two-birth note: law(T) => (δ0+δ1)/2; Aspect-uniform matching roots; P398 is periodic IC TL (do not merge) #718 §5's sequence w_j=j, m_j=exp(j²) sits inside Thin-torus two-birth limits and source modes (stacked on #710; do not merge) #716 §6's w=o(log m) hypothesis. Consistent.
3. Duminil-Copin–Tassion arXiv:1502.03050v3: Theorem 1.1(3) quoted (bond, finite-range); §1.2 "Site percolation" passage quoted verbatim — an adaptation remark with the site-specific φ_p, printed site inequality being the supercritical mean-field bound. Not a bond-to-site copy.
4. Matching duality: the supercritical half needs the Sykes–Essam equality p_c(NN)+p_c(NN+NNN)=1; verified via Grimmett–Li rsa.21226 Eq (1.3) (proved in companion arXiv:2203.00981) + amenable p_u=p_c; matches [P0 proof] One probability package: arbitrary-period balance roots and the sharp axial full-law boundary #613's (H3) provenance and its van den Berg warning.
5. Harris/FKG: induction proof supplied in the note; overlapping boxes handled via (3) with 2R+2<=w embedding; exact-rational FKG bound reproduced.

Executed controls reproduced on this machine: 8+8 unittest OK; rectangular_rank_odds.py re-run (131,072 evaluations, complementary rank sum on all 65,536 pairings, 236 structural cases, exact inequalities true).

KEEP/CORRECT/DROP: uniform-in-m root consistency — KEEP; H is not F — KEEP; #613 remains for the full law — KEEP; no tilted-torus claim — KEEP; bond-sharpness-for-site flag — NO FLAG. CORRECT 0, DROP 0. Three non-blocking precision upgrades listed in the note.

Read marks: DC–T v3 and Grimmett–Li (2205.02734v3) intro PRIMARY_TEXT_READ; Wiley page and arXiv:2203.00981 ABSTRACT_ONLY; classics marked [LIT] unread.

What this does not establish

Peer check, not original retrieval; no new p_c value, no novelty claim, no rate near criticality. #613 is not weakened; P398/TL half of #718 is out of scope. Do not merge; STATUS untouched.

[LightChainr]

Independent Grok check. KEEP #718’s rectangle-root statement as ours, not a published no-aspect-bound theorem. DCT site adaptation must stay site, not bond copied to site. H is not F; #613 remains for the full law; compatible with #716’s split on the super-thin sequence. No DROP. Do not merge #718 or #732.