diff --git a/notes/venue-retrieval-p3-20260909.md b/notes/venue-retrieval-p3-20260909.md
new file mode 100644
index 00000000..ea9f09ad
--- /dev/null
+++ b/notes/venue-retrieval-p3-20260909.md
@@ -0,0 +1,434 @@
+# Venue and literature retrieval for the P3 draft ("Test the ray, not the ratio")
+
+**Date:** 2026-09-09
+**Ticket:** #695 (parent #650; manuscript ticket #579)
+**Kind:** retrieval and notes only. No Monte Carlo, no rescoring, no Huawei, no edits to the
+manuscript's scientific claims, no `docs/STATUS.md`.
+**Method:** primary sources were fetched over the web on 2026-09-09. Each source is marked
+`PRIMARY_TEXT_READ` (full text or page fetched and the quoted sentence read in it),
+`ABSTRACT_ONLY` (only the abstract fetched), `[LIT]` (not fetched; known from a secondary
+source's reference list), or `API_QUERY` (search-API hit counts, not full text). No quotation
+below is from memory: every quoted string was read in the fetched document named beside it.
+
+---
+
+## 1. What the draft claims (the target of this retrieval)
+
+§8 of `docs/manuscripts/p3-projective-inference/manuscript.md` states:
+
+> "The practice of testing amplitude ratios against predicted ratios is ubiquitous — it is how
+> conformal-invariance predictions for aspect-ratio and boundary-condition dependence are compared
+> with lattice data [6] **[LIT]**"
+
+and reference [6] is the draft's only marked hole: Cardy 1984–1986, "specific citation to be
+fixed against the primary." The ticket asks whether primary literature actually **ratio-tests**
+finite-size amplitudes (forms `Y/X` from measured amplitudes and tests it against a predicted
+ratio), and whether the word "ubiquitous" survives contact with that literature.
+
+## 2. Q1 — papers that form a ratio of finite-size amplitudes and test it (the [6] hole)
+
+**Headline result: the practice is real, but it is not "ubiquitous", and the draft's sentence
+must be weakened** (§6 below). The dominant framing in the crossing-probability literature is a
+**full-curve comparison** of the measured probability against the conformal formula as a function
+of aspect ratio; explicit ratio-of-amplitudes tests exist but are the exception, and most sit in
+the universal-amplitude-ratio literature rather than the aspect-ratio literature.
+
+### 2.1 Ratio tests, primary text read
+
+**R1. Simmons, Kleban & Ziff (2007), "Percolation crossing formulae and conformal field theory",
+*J. Phys. A: Math. Theor.* **40**, F771. `PRIMARY_TEXT_READ`**
+(ar5iv full text of arXiv:0705.1933; venue confirmed on the IOPscience page
+https://iopscience.iop.org/article/10.1088/1751-8113/40/31/F03).
+
+The cleanest genuine ratio test found: a measured fraction of Monte Carlo events is compared
+against a predicted **ratio of two integrated conformal densities**. Verbatim, from the
+numerical-verification section:
+
+> "The above fraction of multiple crossing events, 0.07018, is somewhat above the predicted
+> value 0.069189, which is found by integrating the formulas for ν_h and π̄ᵇ_h and taking the
+> ratio of the integral of the former to the sum of the integrals of the former and latter."
+
+and, on the finite-size extrapolation of that ratio:
+
+> "The data fit very well to a straight line when plotted as a function of 1/L, with an intercept
+> of 0.06928, quite close to the predicted value."
+
+Measured `Y/(X+Y)` vs predicted ratio, extrapolated to the infinite-lattice limit — a ratio test
+in exactly the sense the draft describes. It is also honest evidence for the counter-claim below,
+because the same paper's main test is a full-curve comparison ("The relative difference between
+the two curves is on the order of 2%…").
+
+**R2. Izmailian & Yeh (2009), "Ising model with mixed boundary conditions: universal amplitude
+ratios", *Nucl. Phys. B* **814**, 573. `PRIMARY_TEXT_READ`**
+(ar5iv full text of arXiv:1005.1712; journal reference read on the arXiv abs page).
+
+Boundary-condition amplitude ratios, formed and tested against a conformal prediction. Verbatim:
+
+> "In this paper we present the new set of the universal amplitude ratios for the mixed boundary
+> universality class. The results are in perfect agreement with a perturbated conformal field
+> theory scenario proposed by Cardy [cardy86]."
+
+The ratios are of the form `r_s(k) = b_k/a_k` (correlation-length over free-energy first-order
+amplitudes): a ratio of two finite-size amplitudes with different boundary conditions, compared
+with the ratio predicted by perturbed CFT. Caveat stated honestly: these amplitudes come from
+the **exact solution** (Euler–Maclaurin of exact eigenvalues), not from Monte Carlo measurement.
+
+**R3. Kamieniarz & Blöte (1993), "Universal ratio of magnetization moments in two-dimensional
+Ising models", *J. Phys. A* **26**, 201. `ABSTRACT_ONLY`**
+(IOPscience abstract page
+https://iopscience.iop.org/article/10.1088/0305-4470/26/2/009/meta; full text paywalled).
+
+Verbatim from the abstract:
+
+> "The authors calculate the universal critical-point ratios of the square of the second and the
+> fourth moment of the magnetization for ferromagnetic Ising models on the square and on the
+> triangular lattices. … This analysis is also applied to rectangular systems with arbitrary
+> aspect ratios."
+
+A ratio of two finite-size amplitudes from transfer-matrix finite-size data, including
+aspect-ratio-dependent systems, against a universal value. Because only the abstract was read,
+this citation cannot supply a verbatim `Y/X` sentence; it is included as ABSTRACT_ONLY support,
+not as one of the three load-bearing quotations.
+
+### 2.2 Primary papers that compare with lattice data but do **not** ratio-test
+
+These carry the "conformal-invariance predictions … are compared with lattice data" half of the
+draft's sentence — and they show that the comparison is normally a curve, not a ratio.
+
+**C1. Langlands, Pouliot & Saint-Aubin (1994), "Conformal invariance in two-dimensional
+percolation", *Bull. Amer. Math. Soc.* **30**, 1–61. `PRIMARY_TEXT_READ`**
+(PDF fetched from https://publications.ias.edu/sites/default/files/conf-invar-ps.pdf).
+
+Their §3.2 is literally titled "Experimental verification of Cardy's formula." Verbatim:
+
+> "The goal of the first experiment is twofold: to verify again Cardy's prediction for the
+> function π_h(r) on M0 and to obtain values of π_hv(r, M0) suitable for comparison in other
+> experiments."
+
+> "Figure 3.2. Comparison of 81 measured values of ln π_h/(1 − π_h) (dots) with Cardy's
+> prediction (curve)."
+
+> "Despite the systematic error, the agreement is remarkable and we shall compare the results of
+> the following experiments with π^cft instead of π̂ when the former is applicable."
+
+The comparison is of the measured function of aspect ratio against the predicted function; the
+closest thing to a ratio is a log-odds transform of a single probability, not a ratio of two
+amplitudes.
+
+**C2. Watts (1996), "A crossing probability for critical percolation in two dimensions",
+*J. Phys. A* **29**, L363. `PRIMARY_TEXT_READ`** (arXiv HTML of cond-mat/9603167).
+
+Watts compares his formula for π_hv against the Langlands *et al.* Monte Carlo data. Verbatim:
+
+> "Comparing this function with the numerical data for π_hv presented in [2], our agreement is as
+> good as that of Cardy's formula for π_h."
+
+> "In figure 1 we plot log(π_hv) against log(r) for both the numerical results obtained in [2]
+> and for the function 𝓕(x) above. The agreement is excellent."
+
+Full-curve comparison again. Watts also records the historical fact that Cardy's fit to LPS's
+data was so good that the numericians adopted it in place of their own data — which cuts against
+any claim that ratio-testing is the routine quantitative step:
+
+> "In fact, this fit was so good, that Langlands et al. decided to adopt Cardy's result as π_h,
+> rather than their own numerical data, in their later paper on more complicated calculations [2]."
+
+**C3. Ziff (1995), "On Cardy's formula for the critical crossing probability in 2D percolation",
+*J. Phys. A* **28**, 6479. `ABSTRACT_ONLY`** (IOPscience page
+https://iopscience.iop.org/article/10.1088/0305-4470/28/5/013; DeepBlue full text blocked by
+Cloudflare, IOP full text paywalled). The abstract establishes that the paper expands Cardy's
+formula as a function of aspect ratio; the full text could not be read, so it is used here
+neither for nor against.
+
+**[LIT] Langlands, Pichet, Pouliot & Saint-Aubin (1992), "On the universality of crossing
+probabilities in two-dimensional percolation", *J. Stat. Phys.* **67**, 553** — known here from
+Watts (1996)'s reference list [2]; not fetched. The multi-lattice-family universality test and
+the natural next primary to read if [6] is rebuilt.
+
+**[LIT] Pinson (1994), "Critical percolation on the torus", *J. Stat. Phys.* **75**, 1167** —
+the torus-amplitudes companion to LPS. Springer's CDN blocked both PDF and HTML fetches from
+this machine on 2026-09-09 (HTTP 406/403, repeated), and Semantic Scholar records
+`openAccessPdf: CLOSED`. Not read; listed so the reader knows it was attempted and is the next
+thing to fetch for the boundary-condition half of [6].
+
+### 2.3 Verdict on the [6] hole
+
+The ticket asked for three primary papers that *actually ratio-test*. Honest scoring:
+
+| candidate | ratio of measured amplitudes vs predicted ratio? | status |
+|---|---|---|
+| Simmons–Kleban–Ziff 2007 | **yes**, explicitly (measured event fraction vs predicted ratio of integrated densities, extrapolated in 1/L) | PRIMARY_TEXT_READ |
+| Izmailian–Yeh 2009 | **yes**, but the amplitudes are exact-solution, not simulated | PRIMARY_TEXT_READ |
+| Kamieniarz–Blöte 1993 | **yes** (universal ratio from finite-size data incl. aspect ratios), but abstract only | ABSTRACT_ONLY |
+| LPS 1994; Watts 1996; Ziff 1995 | no — full-curve comparisons | PRIMARY_TEXT_READ ×2, ABSTRACT_ONLY ×1 |
+
+Three papers can be cited — but the third only on its abstract, and two of the three are not
+aspect-ratio ladder measurements in the draft's sense. What the primary reading actually shows
+is that the **routine** comparison in this literature is the full curve of a probability or
+amplitude against its conformal prediction as a function of aspect ratio (LPS 1994; Watts 1996),
+with ratio-of-amplitudes tests appearing as secondary or special-purpose checks. The draft's
+"ubiquitous" therefore does not survive, and §8 must be weakened as in §6.
+
+## 3. Q2 — is the projective statistic new?
+
+**Bottom line: the exact statistic as a *test* appears new to the FSS/percolation literature,
+but narrowly, and the draft must engage four pieces of nearby prior art it does not currently
+cite.**
+
+**Negative evidence (novelty survives):**
+
+- Inspire HEP API, query `Fieller confidence interval`: **0 hits**. `API_QUERY`
+- arXiv metadata search, query "Fieller": 4 hits, all statistics — Owen (2025,
+  arXiv:2510.00389), Tang (2021, arXiv:2110.12636), von Luxburg & Franz (2009, arXiv:0711.0198),
+  Franz (2007, arXiv:0710.2024). None in FSS/percolation/stat-mech. `API_QUERY` /
+  `ABSTRACT_ONLY`
+- "Fieller" + "scaling amplitude": 0 hits; "Hotelling" + "finite-size scaling": 0 hits;
+  "finite-size scaling" + covariance + amplitude fits: 0 hits. `API_QUERY` (arXiv search is
+  metadata-only, so absence is weak-but-indicative)
+- Gleser–Hwang (1987): only statistics sources found; no physics citation evidence. Negative
+  result.
+- `pyfssa` documentation, https://pyfssa.readthedocs.io/en/stable/reference/fssa.html.
+  `PRIMARY_TEXT_READ`. The standard FSS data-collapse tool's quality statistic is
+  covariance-free: "This is the reduced chi-square statistic for a data fit except that the
+  master curve is fitted from the data itself" — its parameters are `x, y, dy` per-point
+  standard errors, with no covariance input (quality function from Houdayer & Hartmann,
+  *Phys. Rev. B* **70**, 014418 (2004)).
+
+**Prior art that must be cited and distinguished:**
+
+**P1. Parisen Toldin (2011), "Improvement of Monte Carlo estimates with covariance-optimized
+finite-size scaling at fixed phenomenological coupling", *Phys. Rev. E* **84**, 025703(R).
+`PRIMARY_TEXT_READ`** (arXiv abs + HTML of arXiv:1104.2500; quotes independently re-verified
+directly in this retrieval). Verbatim (abstract):
+
+> "Within this scheme of finite-size scaling, we exploit the statistical covariance between the
+> observables in a Monte Carlo simulation in order to reduce the statistical errors of the
+> quantities involved in the computation of the critical exponents."
+
+Verbatim (method section; the exact phrase "orthogonal projection" does not appear, but the
+orthogonal/parallel decomposition does):
+
+> "A positive-definite scalar product <,> between two vectors δ and δB̂ can be defined by the
+> covariance <δÂ, δB̂> ≡ COV[δÂ, δB̂] = E[δ δB̂] − E[δÂ] E[δB̂]"
+
+with δO decomposed into components "orthogonal δO⊥ and parallel δO∥ to the space 𝒱". This is the
+covariance-metric geometry of the draft's statistic, already inside FSS — but it *minimizes the
+variance of an error-propagated observable* (norm minimization under the covariance metric,
+projection onto an affine subspace of dimension N−1, not a ray): no χ², no hypothesis test, no
+null inversion. The draft's claim survives as "the projective statistic as a *test statistic*
+is new", and §1.4/§8 should cite Parisen Toldin (2011) as the nearest neighbour. The abstract's
+"the natural statistic is then…" should acknowledge that a covariance scalar product is already
+in FSS use for error minimization.
+
+**P2. von Luxburg & Franz (2009), "A Geometric Approach to Confidence Sets for Ratios: Fieller's
+Theorem, Generalizations, and Bootstrap", *Statistica Sinica* **19**, 1095; arXiv:0711.0198.
+`ABSTRACT_ONLY`.** A geometric generalization of Fieller in the statistics literature; the
+draft's §2.4 (confidence regions for directions) and §1.4 ("it is not new mathematics") should
+name it, because it is the statistics-side generalization closest to what the draft builds.
+
+**P3. Bruno & Sommer (2023), "On fits to correlated and auto-correlated data", *Computer Physics
+Communications* **285**, 108643. `PRIMARY_TEXT_READ`** (abstract + introduction + conclusions
+fetched; PDF paywalled). Verbatim (abstract):
+
+> "Standard χ2 tests require a reliable determination of the covariance matrix and its inverse
+> from correlated and auto-correlated data, a challenging task often leading to close-to-singular
+> estimates. These motivate modifications of the definition of χ2 such as uncorrelated fits."
+
+Correlated-χ² fitting is standard practice in lattice QCD; this supports the draft's framing of
+the pseudo-inverse/rank-reporting discipline (§2.2) rather than competing with it.
+
+**P4. Wang, Iyer & Hannig (2003), "Uncertainty Calculation for the Ratio of Dependent
+Measurements", *Metrologia*. `PRIMARY_TEXT_READ`** (NIST publication page). Verbatim (abstract):
+
+> "Although an exact confidence interval procedure, known as the Fieller interval, is available
+> for this problem, practitioners often use various non-exact methods."
+
+Fieller is alive in metrology but absent from physics FSS — which is exactly the draft's premise,
+now documented rather than assumed.
+
+**Standard-practice control.** Mertens, Jensen & Ziff (2017), "Universal features of cluster
+numbers in percolation", *Phys. Rev. E* **96**, 052119; arXiv:1602.00644. `PRIMARY_TEXT_READ`.
+Verbatim: "The metric factor b cancels out in the dimensionless ratio R = A₁C₁/B₁² =
+f(0)f″(0)/2f′(0)² which is predicted to be universal for systems of a given shape." The
+dimensionless-ratio practice the draft criticizes, in the percolation literature, formed by
+cancelling the denominator — no covariance, no ray test anywhere in the fetched text.
+
+## 4. Q3 — annihilating functionals / divided differences as class-level tests
+
+**Bottom line: honest null.** Searches run: OpenAlex fulltext "divided differences" +
+"finite-size scaling" (4 hits, all quantum Monte Carlo papers with no FSS-amplitude content);
+arXiv metadata "divided differences" + "scaling amplitudes" (0 hits); web searches for
+divided-difference or linear-combination class tests of critical/scaling amplitudes (only
+generic scaling reviews and Kenna-type hatted-exponent amplitude **relations**, which are built
+from ratios, not from a functional annihilating a class); "divided differences" lattice operator
+(numerical-analysis and lattice-quasi-PDF usages only). `API_QUERY` and search-result level —
+the only `[LIT]`-grade items here are the Kenna-type amplitude-relation papers
+(arXiv:2402.00427 / *Entropy* **26**, 221 (2024)), which are ratio-based, not annihilator-based.
+
+The closest construct in the fetched literature is Izmailian & Yeh (2009)'s universal linear
+combinations `r_s(k) = b_k/a_k` — but those test individual predicted ratios, not a functional
+that annihilates an entire class whatever the amplitudes. The draft's §8 sentence "we have not
+seen it deployed as a class-level tool for finite-size ladders" stands as phrased ("we have not
+seen"), and should not be strengthened to a proven negative: Google Scholar and the paywalled
+PRB/PRE archives were not fully searchable.
+
+## 5. Q4/Q5 — venue fit and the one-dataset reanalysis precedent
+
+### 5.1 Q4: journals (scope sentences verbatim from the publisher pages named)
+
+**Physical Review E — FIT.** Scope (journals.aps.org/pre/about):
+
+> "Physical Review E is a trusted and interdisciplinary journal for high-quality, significant
+> developments in the interrelated areas of statistical, nonlinear, complex systems, biological,
+> soft matter, fluids, plasma and computational physics, including machine learning and
+> artificial intelligence."
+
+Acceptance criterion (same page): "Add to the existing knowledge related to statistical,
+nonlinear, biological, and soft matter physics." A covariance-weighted χ² test of FSS amplitude
+vectors in lattice percolation sits in PRE's statistical/computational core, and PRE explicitly
+publishes reanalyses (see §5.2). **Analogue:** Jensen & Ziff, "Universal amplitude ratio Γ−/Γ+
+for two-dimensional percolation", *Phys. Rev. E* **74**, 020101(R) (2006) — an amplitude-ratio
+determination paper in PRE (`ABSTRACT_ONLY`, arXiv:cond-mat/0607146).
+
+**JSTAT — FIT.** Scope (IOP publishing-support page, the direct /about page being blocked):
+
+> "JSTAT is targeted to a broad community interested in different aspects of statistical
+> physics, which are roughly defined by the fields represented in the conferences called
+> 'Statistical Physics'."
+
+Topically exact. Closest verified analogue: Xun *et al.*, *J. Stat. Mech.* (2025) 123301
+(extended-range percolation, extensive MC + FSS); **no true amplitude-ratio-testing JSTAT
+analogue was located — noted honestly.**
+
+**J. Stat. Phys. — MILD STRETCH.** Springer's aims-and-scope page was bot-blocked; the only
+description obtained is Wikipedia's ("All areas of statistical physics as well as related
+fields concerned with collective phenomena in physical systems are covered"), which is **not a
+publisher quote**. Topically a fit, but JSP's culture favours formal theory over a
+Monte-Carlo-reanalysis methodology paper.
+
+**Computer Physics Communications — STRETCH.** Scope (sciencedirect.com
+aims-and-scope): "Computer Physics Communications publishes research papers and application
+software in the broad field of computational physics" … "The focus of CPC is on contemporary
+computational methods and techniques and their implementation, the effectiveness of which will
+normally be evidenced by the author(s) within the context of a substantive problem in physics."
+Acceptable only if the method ships as software with the paper. Analogue: Hu, Chen, Izmailian &
+Kleban, *CPC* **126**, 77–81 (2000) (Monte Carlo methods for percolation).
+
+**The American Statistician — STRETCH.** Scope (journal landing page): "articles that address
+practical problems at the interface between statistical methodology and areas of application."
+The generalized-Fieller statistic fits that brief, but the physics framing must lead or the
+audience match fails; and the paper has no simulation study, so a statistics journal would
+treat it as a note. Analogue: Hirschberg & Lye, "A Geometric Comparison of the Delta and
+Fieller Confidence Intervals", *The American Statistician* **64**(3), 234–241 (2010).
+
+**J. Phys. A — FIT.** Scope (IOP publishing-support page): acceptable papers must contribute to
+listed sections, which include verbatim "statistical mechanics, phase transitions and critical
+phenomena" and "numerical and computational methods, analysis of algorithms". Analogue: Delfino,
+Viti & Cardy, "Universal amplitude ratios of two-dimensional percolation from field theory",
+*J. Phys. A* **43**, 152001 (2010).
+
+### 5.2 Q5: does PRE accept a reanalysis of one dataset that inverts a verdict?
+
+**Yes — precedent exists.** Three fetched examples:
+
+1. **Wang, Zhou, Zhang, Garoni & Deng (2013), "Bond and site percolation in three dimensions",
+   *Phys. Rev. E* **87**, 052107. `PRIMARY_TEXT_READ`** (full text via harvest.aps.org).
+   Revises the long-standing simple-cubic bond threshold 0.248 812 6(5) down to
+   0.248 811 82(10). Verbatim: "Figure 1 shows that our estimate of pc lies slightly below the
+   central value 0.248 812 6 reported in Ref. [21]."
+2. **Mertens (2022), "Exact Percolation Probability on the Square Lattice", *J. Phys. A*
+   **55**. `PRIMARY_TEXT_READ`** (ar5iv of arXiv:2109.12102). Weakens every prior Monte Carlo
+   estimate of the square-lattice site threshold. Verbatim (abstract): "We use the data to
+   compute estimates for the percolation threshold pc that are several orders of magnitude more
+   precise than estimates based on Monte-Carlo simulations." Verbatim (text): "The first value
+   deviates from the reference value [19] by 2 errorbars, the second by 2.5 errorbars." And:
+   "as Bob Ziff put it, 'Monte-Carlo is dead' [40] when it comes to computing pc for 2d
+   systems."
+3. **Jensen & Ziff (2006), *Phys. Rev. E* **74**, 020101(R). `ABSTRACT_ONLY`.** Fixes a
+   previously scattered verdict on a percolation amplitude ratio. Verbatim (abstract):
+   "Putting all methods together we find a consistent value Γ−/Γ+=162.5±2, a significant
+   improvement over previous results that placed the value of this ratio variously in the range
+   of 14 to 220."
+
+So the draft's Q5 worry — that PRE would not take a methods paper whose worked example cannot
+name a surviving law — is answerable from these: PRE has published reanalyses whose whole point
+was that a previously accepted number or verdict was wrong, including one (Jensen–Ziff) whose
+result is precisely the fixing of a scattered amplitude-ratio verdict. The draft's undetermined
+eighth verdict, honestly reported as undetermined, is within that precedent.
+
+## 6. Stress-test of the prior Grok assessment
+
+1. **"PRE is the right first venue; the amplitude-ladder literature lives there" — AGREE.**
+   PRE's scope sentence (§5.1) plus the fetched PRE record: Jensen–Ziff 2006 (amplitude ratio),
+   Mertens–Jensen–Ziff 2017 (universal ratio of cluster-number amplitudes), Wang *et al.* 2013
+   (threshold reanalysis). J. Phys. A is a defensible second; JSTAT third.
+2. **"The math is not new; novelty is diagnosing the design pathology" — AGREE, with one
+   refinement.** The math is not new, but the retrieval shows the novelty claim must be narrowed
+   and cited, not just asserted: Parisen Toldin (2011) already uses the covariance scalar
+   product and an orthogonal decomposition inside FSS (for error optimization), and von
+   Luxburg–Franz (2009) already generalize Fieller geometrically (in statistics). The novel
+   content is the *test statistic with a χ² tail on rank(S) − dim(V) d.o.f.*, the diagnosis of
+   the self-sealing denominator choice, and the worked inversion — not the geometry.
+3. **"Without 3–4 primary ratio-testing citations the literature-practice claim dies" —
+   PARTIALLY CONFIRM, and go further.** Three primaries exist (§2.1), so the practice is not
+   undocumented; but the primary reading shows the *dominant* practice is full-curve comparison,
+   so "ubiquitous" dies even with the citations in hand. Weaken the sentence (§6 of the
+   manuscript changes below) rather than propping it up with the three papers.
+4. **"Statistics journals will call this a note" — AGREE.** TAS's own scope ("practical
+   problems at the interface…") plus the absence of any simulation study makes a statistics
+   venue a stretch; Hirschberg & Lye (2010) is the analogue shape there.
+5. **"Do not claim a percolation-threshold or modular-weight result" — AGREE, consistent with
+   the draft's own §5.3.** Nothing in this retrieval touches that boundary.
+
+## 7. What this retrieval supports changing in the draft
+
+**The one sentence that must change** is §8's
+
+> "The practice of testing amplitude ratios against predicted ratios is ubiquitous — it is how
+> conformal-invariance predictions for aspect-ratio and boundary-condition dependence are compared
+> with lattice data [6]"
+
+Proposed replacement, consistent with what was actually read:
+
+> "Conformal-invariance predictions for aspect-ratio and boundary-condition dependence are
+> routinely compared with lattice data, usually by full-curve comparison of the measured
+> probability against the predicted one (Langlands, Pouliot & Saint-Aubin 1994; Watts 1996);
+> explicit tests of a ratio of measured amplitudes against a predicted ratio exist (Simmons,
+> Kleban & Ziff 2007; Izmailian & Yeh 2009; Kamieniarz & Blöte 1993) but are the exception rather
+> than the routine form of the comparison."
+
+Proposed bibliographic replacement for [6] (quotes in §2):
+
+- [6a] Cardy, J. L. (1992), *J. Phys. A* **25**, L201 — the theory source the sentence wants. `[LIT]`
+- [6b] Langlands, Pouliot & Saint-Aubin (1994), *Bull. Amer. Math. Soc.* **30**, 1–61. `PRIMARY_TEXT_READ`
+- [6c] Watts, G. M. T. (1996), *J. Phys. A* **29**, L363. `PRIMARY_TEXT_READ`
+- [6d] Simmons, Kleban & Ziff (2007), *J. Phys. A* **40**, F771. `PRIMARY_TEXT_READ`
+- [6e] Izmailian & Yeh (2009), *Nucl. Phys. B* **814**, 573. `PRIMARY_TEXT_READ`
+- [6f] Kamieniarz & Blöte (1993), *J. Phys. A* **26**, 201. `ABSTRACT_ONLY`
+
+Additional citation debts found (for the authors' consideration, not patchable here):
+
+- **Parisen Toldin (2011)**, PRE **84**, 025703(R), in §1.4/§8 — nearest prior use of the
+  covariance-metric geometry in FSS (error optimization, not testing). `PRIMARY_TEXT_READ`
+- **von Luxburg & Franz (2009)**, *Statistica Sinica* **19**, 1095, in §2.4 — the statistics-side
+  geometric generalization of Fieller. `ABSTRACT_ONLY`
+
+Per the ticket's tripwire: this note does **not** repeat the "ubiquitous" claim, and the "we
+correct a literature practice" framing should be softened to "we recommend a statistic whose
+two-entry case is the known-correct special case of the ratio practice" unless the full texts of
+Pinson (1994) and Langlands *et al.* (1992) turn up further ratio-testing on reading.
+
+## 8. Sources attempted and blocked
+
+- Springer CDN (Pinson 1994 PDF/HTML; JSP aims-and-scope): HTTP 406/403 IP-block and bot
+  challenge, 2026-09-09.
+- University of Michigan DeepBlue (Ziff 1995 full text): Cloudflare interstitial, twice.
+- ADS (Pinson abstract): human-verification interstitial.
+- journals.aps.org direct article pages: blocked/404; PRE full text obtained via
+  harvest.aps.org.
+- IOPscience /about pages for JSTAT and J. Phys. A: direct fetch failed; scope quotes taken
+  from IOP's official publishingsupport.iopscience.iop.org pages instead.
+- APS (a guessed Kamieniarz–Blöte DOI page): 404; the IOP abstract page was used instead.
+- Reachable and used: arXiv (abs/HTML), ar5iv, IOPscience abstract pages, IAS publications,
+  KCL Pure, harvest.aps.org, Inspire HEP API, OpenAlex API, Semantic Scholar API, NIST
+  publication pages, pyfssa docs, publisher aims-and-scope pages for Elsevier and TAS.