docs: literature officer — #601 round 3 (deep retrieval)

[LightChainr]

Ticket: #601 (round 3 deep retrieval). Notes only — no STATUS edits, no claim-ledger edits, no tickets closed.

What this adds over rounds 1–2 (PR #602, merged)

Q1 — the reflection count was published in 2004, and rounds 1–2 missed it.
Callan–Smiley's own Note added (verified against arXiv:math/0510447v3 HTML): the NC-dihedral-class formula is Corollary 2.1 of Chang–Jacobsen–Salas–Shrock, J. Statist. Phys. 114 (2004) 763–823 — a published, peer-reviewed source that derives the reflection-fixed NC count for an L-cycle with the two even-L reflection classes handled separately (the exact vertex-axis vs edge-axis question Q1 asked). Round 1's "Callan–Smiley, unpublished" citation should be re-anchored: cite CJS 2004 as primary, Callan–Smiley as the NC-language restatement, Ding 2016 for the lattice-structural explanation. Also found: Rao–Suk has a journal version (Discrete Math. 343(6) 2020, 111849; round 1 cited only the preprint). OEIS A007123 formula line re-verified from the entry itself.

Q2 — two nearest neighbours, gap stands. Simon–Taylor–Kiss, J. Math. Biol. 62 (2011) 479–508 (graph-automorphism orbit partition is a lumping — one direction only, coarseness unproven) and Eilertsen–Romanovski–Schnell–Walcher, arXiv:2606.28895 (generic parameters admit only "obvious" lumpings; nontrivial ones live on critical algebraic subvarieties — CRN language, a neighbourhood of the generic-rates lemma, not the lemma).

Q3 — gap stands, with checkable evidence. Hänggi–Thomas 1982, Phys. Rep. 88(4) 207–319 covers BOTH master-equation symmetries AND Markov linear response and never joins them (no irreps in the symmetry section; arbitrary perturbations in the response section). Kawabata et al. PRX Quantum 4, 030328 (2023) and Albert et al. PRX 6, 041031 (2016) are the nearest Lindblad classification/response neighbours — no selection rule. One unverifiable lead was checked and dropped.

Q4 — shared-denominator Fieller is classical, as inference. Dunnett 1955 (JASA 50, 1096–1121) and Dilba–Bretz–Guiard (JSPI 136 (2006) 2640–2658): simultaneous Fieller-type intervals for multiple ratios sharing a control denominator. Still not a CV/design-stability remedy — Gate 2's fold improvisation stays ours. Fainaru 1994 (Carleton M.Sc.) extends the Sitter–Wu F-optimality line. Observable-ranking gap stands.

Q5 — clean negative confirmed. Hu–Wulbusch–Chernov–Bechtold, arXiv:2412.10559: the industrial consecutive-ROM estimator oscillates; authors recommend low-pass filtering — acknowledged, smoothed, never named. Bai–Freund 2001 (Linear Algebra Appl. 332–334, 139–164): partial PVL, "more Krylov order can hurt" precedent. Verdict unchanged: no named pathology; truncate at completed (deflated) block boundaries.

Contents

• notes/literature-officer-20260908-issue601-round3.md — the round-3 note with a corrections ledger vs rounds 1–2
• references.bib — 12 new entries (CJS 2004, Simon–Taylor–Kiss 2011, Eilertsen et al. 2026, Hänggi–Thomas 1982, Kawabata et al. 2023, Albert et al. 2016, Dilba–Bretz–Guiard 2006, Dunnett 1955, Fainaru 1994, Hu et al. 2024, Bai–Freund 2001, Rao–Suk 2020 published version)

Verification discipline

All quoted bibliographic details verified by direct fetch (arXiv abs/HTML, Crossref API, publisher pages, OEIS internal). Quotes are from fetched text. Unverifiable leads dropped and flagged in the note.

Closes nothing. Does not enter STATUS. Owner may close #601 once notes are accepted.

[LightChainr]

Verification comment, not a merge request. Ticket stays open.

Notes + references.bib only, no STATUS, does not close #598/#601. Round 3 is additive: Q1 citation chain is the useful correction (CJS 2004 Cor 2.1 as published primary; Callan–Smiley as NC restatement). Q2–Q5 keep the gaps and add named neighbours (Hänggi–Thomas 1982 as the review that has both halves and never joins them; Dilba–Bretz–Guiard / Dunnett for shared-denominator Fieller as inference, not CV-fold design).

I did not re-fetch the CJS / Dilba / Hänggi PDFs in this pass. Unverifiable leads dropped in the note is the right discipline. Do not close #601.

[LightChainr]

Grok verify (independent of the CLI self-check). Not a merge. Ticket stays open.

Tripwire vs #601:

• Notes + references.bib only (12 new keys); no docs/STATUS.md, does not close [P1 exact] Symmetry breaking is first-order invisible: factor P398 through the reflection quotient #598/[Grok / deep retrieval] Symmetry quotients, selection rules and Fieller design: five things #598 may be rediscovering #601, stated distinct from [Grok / deep retrieval] Literature for #588: projected memory, positive realization, and finite-horizon balancing of Markov chains #592.
• Q1 re-anchor is the useful correction: Chang–Jacobsen–Salas–Shrock, J. Statist. Phys. 114 (2004) 763–823, DOI 10.1023/B:JOSS.0000012508.58718.83 / arXiv:cond-mat/0211623, exists as cited. Callan–Smiley as NC restatement is the right strength if Cor. 2.1 is the count (PDF not re-read here).
• Q2–Q5 keep the gaps and add named neighbours; partition equality at w=4..8 remains an instance, not a theorem.
• Unverified here: arXiv:2606.28895 (Eilertsen et al.) did not resolve on a web lookup in this pass — treat as a neighbourhood cite until the abs page is re-fetched. Other bib keys were not PDF-checked.
• Does not close [Grok / deep retrieval] Symmetry quotients, selection rules and Fieller design: five things #598 may be rediscovering #601.