docs(140): §2ter — #250 Rules + #435 Mémo résolus sur branches (validation guide)

[jsboige]

Objet

Met à jour le guide de validation #140 (docs/publication/validation-guide.fr.md) pour refléter que les 7 problèmes #250 (Rules) + #435 (Mémo) sont résolus sur branches et validés visuellement ce cycle. Le §2bis existant décrivait encore l'état master/Release pré-fix (Rules 24 cartes « pending re-check jusqu'au fix #250 », Mémo absent du harnais) — ce qui aurait induit en erreur lors du sign-off #140.

Changement

• Header : note §2ter (2026-06-04).
• Nouveau §2ter : statut fix(rules): Rules cards text spacing and layout issues #250 (fix(rules): editorial restructure Tarot Rules 24→15 cards + 15-card CSS remap (#250, supersedes #437) #438, 24→15 cartes, 0px overflow, miroir 8 langues byte-identique) + fix(rules): Mémo cards — problèmes non corrigés signalés par jsboige #435 (fix(memo): sélecteur de famille language-invariant — Mémo dans toutes les langues (#435) #439, sélecteur i18n, 7/7 familles ×4 langues), la question éditoriale Mémo Back (clés taxonomie FR), et la séquence merge→régén Target propre pour basculer le dossier au vert.
• §2bis : la ligne « Rules = pending re-check » est marquée levée (renvoi §2ter).

git diff --stat = 1 fichier, +24/−2. Docs-only.

Statut

🔒 TENUE — contexte pour le sign-off visuel #140 (jeudi). Ne pas merger avant la décision jsboige (référence des PRs #438/#439 encore non mergées).

Refs #140, #250, #435.

[clusterManager-Myia]

[NanoClaw] Review PR #440

Single-file doc update to validation-guide.fr.md (+24/-2). Adds §2ter documenting the resolution status of #250 (Rules restructure 24→15 cards, PR #438) and #435 (Memo i18n selector fix, PR #439), both validated visually on feature branches pre-merge.

Verified:

• All cross-references resolve: #250, #435, PR #438, PR #439 all exist and match described scope
• Section numbering (§2ter) is consistent with existing structure (§1, §2, §2bis, §3…)
• Header date line correctly updated with §2ter entry
• §2bis caveat about pending Rules re-check properly updated to point to §2ter
• Open editorial question (Memo Back FR keys) clearly flagged as a decision item for jsboige
• No security concerns (no secrets, tokens, or credentials)

Observations (non-blocking):

• The §2bis harness data (24 Rules cards, rules_23 Parlote) is accurately flagged as post-merge obsolete — good forward-looking documentation
• Merge sequence (1. #438, 2. #439, 3. regenerate, 4. visual re-check) is clear and actionable
• docs/investigations/ folder references are not linkable in Markdown (no relative path or URL) — consider adding them if those folders will be committed alongside this section

No blocking issues. LGTM for documentation.

[clusterManager-Myia]

[NanoClaw] Re-review PR #440 (head pushed: 7b651ec → 5c01185)

Single new commit on this docs-only PR, correcting a false overflow measurement claim in §2ter.

What changed (commit 5c01185):

• Removed the incorrect "0 px overflow FR (scrollHeight == clientHeight)" claim from the #250 Rules subsection.
• Added a detailed blockquote (⚠️ Correction overflow) documenting why the original measurement was wrong (container auto-fit without bounded height ≠ body-vs-footer collision).
• Inserted corrected before/after overflow data from the live #190 engine: specific per-lang per-card pixel values (worst PT #12 +95px), and confirmation that commit 74ef097 (.texte hook + max-height: 390px, on PR #438 branch) resolved all non-cover collisions across 8 languages.
• Removed the "régén Debug" parenthetical from the validation statement (the overflow fix is engine-side, not a regeneration artifact).
• Commit message accurately describes the correction and scope.

Verified:

• Previous review (#4425425566) was against 7b651ec — this review covers the delta.
• The corrected data is internally consistent: the before numbers (FR/RU/PT collisions, EN clean) match what a bounded-height measurement would produce, and the CSS fix target (card:not([class~="1"]) .texte { max-height: 390px }) plausibly resolves them.
• The explanation of why the harness (#412/#431) missed this (never piloted the live #190 engine) is logically sound and consistent with §2bis's description of harness scope.
• All existing review observations from the first pass remain valid (cross-refs, section numbering, §2bis caveat, editorial question).
• Security scan: no secrets, tokens, or credentials in the full diff.

No blocking issues. The correction is honest and well-documented — exactly the kind of transparent fix you want in a validation guide.