Exact foundations: all-size wrapping labels, corrected shape symmetry, irreducible hidden criticality

[LightChainr]

Direct analysis under the owner's 2026-09-12 delegation. No new production and no new task opened for work completed here.

1. Proves the allowed five-cell wrapping support on every honest torus from the existing digital-Alexander theorem: a line in a symplectic plane is its own orthogonal. Corrects [Superseded by #702] Fact 2 label map — symplectic orthogonality corrected #690 and [Superseded by #702] Wrapping dictionary — false conjecture and obstruction dependencies removed #693; rank-one directional equality and spiral set identity require no extra conjecture or census.
2. Refutes [Superseded by #702] Shape rerun — raw-p C3 withdrawn; valid repaired data retained #666's raw-p C3 as a shape criterion unless the limiting centre is 1/2; gives the correct anchored quantile diagnostic. Also corrects the claimed mismatch of two Aff(1) actions: pushforward and inverse quantile DO intertwine.
3. Replaces the reducible direct-sum witness with an OVERALL IRREDUCIBLE, reversible, bounded-degree product-chain construction with identical task responses but gap-closing locations 1/2 vs 1/3. Abstract Markov control, not percolation; no novelty claim.
4. Narrows [Superseded by #702] Finite-sector obstruction — unrestricted no-go withdrawn #675's blanket one-operator impossibility claim and corrects regular-polygon nonnegative rank to Theta(log n), with a primary text read.

Validation performed locally: 5 targeted checks, including 54 response checks at 40 decimal digits (max diagnostic error 2.30e-41). Full repository suite NOT run in this connector-only environment. All existing production artifacts and historical notes are preserved. Note explicitly identifies which old arguments it supersedes.

References #635, #658, #672, #674, #677, #678, #679, #681, #686, #622, #624, #644, #697. The cylinder/pTL bridge and percolation shape limit remain open. No automatic large-compute release for #636.