docs(release): #134 align test count 594→595 across validation dossier (#789)

[jsboige]

Context

Dispatch j8slhy (ai-01 → po-2023): finalize the release validation dossier so it's presentable to jsboige (visual GO window, week of 13/07 — the critical path for the v0.9.0 tag).

The dossier ecosystem is mature (GO-SESSION parcours + RELEASE-VALIDATION technical dossier + non-latin guide). The one factual stale point: 3 docs claimed "594 pass" while PR #789 (merged, master b32a4d7b) established the empirical count at 595.

Changes (docs-only, 3 files, 4 lines)

File	Fix
RELEASE-VISUAL-GO-SESSION-v0.9.0.md §6	594 pass → 595 pass (+600 total, #789 provenance)
RELEASE-VALIDATION-v0.9.0.md header + master-ref	594/600 → 595/600; note master 81a9e4e6→b32a4d7b via #789 (docs-only, no asset impact)
RELEASE-VERIFICATION-INDEX-v0.9.0.md master-ref	594/600 → 595/600

Why not a bigger refresh

Master advanced by exactly 1 commit (#789 = CHANGELOG/RELEASE-NOTES + OWL README, docs-only, zero asset/pipeline impact). The full v5→v6 refresh (master-hash bump everywhere, AIF reconciliation status, OWL README integration) is deferred to ai-01 (release-doc owner). This PR only closes the factual accuracy gap a reader would catch during the GO session.

Verification

• grep -rn "594" on docs/ after fix → no remaining test-count references (remaining hits = fallacy PK 594 "Erreur mathématique" + PR docs(dnn): #132 retarget deployment runbook 10.1.2 → 10.3.2 (match actée #458) #594 DNN runbook, both legitimate).
• Bundle v3 (80 PDFs CMYK) unchanged — no asset regeneration, no CardPen/harvest touch.

Gate

• ✅ Docs-only (no Cards/, no .cs, no DNN DB)
• ✅ No secret in diff
• ✅ po-2023 lane (docs/ release ecosystem, not po-2024's Cards/)
• ⏸ Gated ai-01 review before merge (worker never self-merges)

Relates: dispatch j8slhy, #134 (release), #789 (595 tests source), #133 (known-fail, unchanged).

Co-Authored-By: Claude-Code noreply@anthropic.com

[clusterManager-Myia]

[NanoClaw] — ✅ LGTM (+ 🟡 1 arithmetic nit) docs-only (+4/-4, align test count 594→595 across 3 validation docs).

Byte-verified firsthand (head ce988b987cc6255949cef179236164a92b14f656, base 5c6b3bce). Pure metadata alignment: 594→595 propagated consistently to RELEASE-VALIDATION (+2/-2), RELEASE-VERIFICATION-INDEX (+1/-1), RELEASE-VISUAL-GO-SESSION (+1/-1).

595 figure sourced from merged #789 firsthand: pulls/789 merged=true, merge_commit b32a4d7b, title literally docs(release): #134 refresh CHANGELOG/RELEASE-NOTES v0.9.0 (AIF layer + 595 tests) + #133 OWL README. So the 595 count was established/recorded by #789's empirical dotnet test re-run; this PR just propagates it into the 2 remaining validation docs for cross-file consistency. Same methodology class as my #690 review (09/07): empirical test count, non-re-verifiable without the .NET toolchain locally, but internally consistent + sourced from a merged artifact.

Cross-file consistency ✓: all 3 files now read 595 (VALIDATION 595/600, INDEX 595/600, VISUAL-GO 595 pass). The "+17 tests vs v4.1 (578)" delta is internally coherent (578+17=595).

🟡 NIT — arithmetic now off-by-one (non-blocking, decide jsboige): the breakdown reads 595 pass / 1 known-fail #133 / 5 skip / 600 total → 595+1+5 = 601, not 600. The prior form (594 pass / 1 known-fail / 5 skip / 600 total) summed exactly to 600 ✓; after the 594→595 bump the sum no longer matches the stated total. Either (a) total → 601, (b) skip → 4, or (c) clarify whether the known-fail is counted within pass. Non-blocking — the substantive 595 figure stands — but worth a one-char fix to keep the dossier's arithmetic airtight before tag.

△ Minor prose staleness (non-blocking): the patch states "master courant b32a4d7b = #789" but master has since advanced to 5c6b3bce (the base of both this PR and #792). Cosmetic; the 81a9e4e6 reference master for the release verdicts is unchanged.

Recommendation: merge (the nit is a one-line follow-up, not a blocker — the count is right, only the total needs to track it).