docs(dnn): refresh 3 #131/#132 runbooks post-execution (10.3.2+2sxc21 delivered)

[jsboige]

Context

Dispatch 2lnzsb (ai-01 → po-2023) [secondaire]: "scope/refresh les 3 runbooks existants (131-regen-staging, 131-step1-sandbox-upgrade, 132-deployment)." The runbooks were written Jun 15-25 as PRE-execution prep — before the migration was delivered. They now describe steps that have already happened, or (in one case) carry an undocumented lesson that would bite a future executor.

What landed (docs-only, 3 files, +85/-1)

Runbook	Refresh
131-step1-sandbox-upgrade-runbook.md	POST-EXEC banner: sandbox upgrade DELIVERED (10.3.2+2sxc21 live full-IIS, ACME bypass active). Points to 131-dnn-phase2-exec-rollback + MEMORY + UPGRADE-ASSESSMENT as current truth. Flags 2 staleness items resolved (§2 "2sxc 15.02" was stale — already 21.07; §5 "templates untouched" moot — cliff crossed).
131-regen-staging-runbook.md	REFRESH banner: scope DEFERRED 4-vs-8 lang → RESOLVED (8-lang shipped #565, 41 SVGs). Foreground-lock mitigated (#569 tscon recipe), Playwright deadlock resolved (#651), build-server deadlock lesson. Points to regen-readiness-refresh-c1ed77d2 (#782) as the live harness.
132-deployment-runbook.md	POST-EXEC banner (sandbox+full-IIS delivered, prod VPS go-live remains gated) + the critical B1-inversion lesson added to §5.5(a).

The one real gap fixed (132 §5.5a)

The deployment runbook predates the B1 incident and would lead a future executor to break 2sxc 21:

The earlier repair-bin-net48.ps1 reverted BCL/SDK DLLs to 6.0.0.0, treating .NET 8/9 versions as "contamination." That thesis was INVERTED for 2sxc 21: 2sxc 21.07 is compiled against System.Text.Json 9.0.0.0 / SCI 9.0.0.0 / Bcl 8.0.0.0. Reverting → JsonOptions type-init MissingMethodException → every 2sxc module "Something went really wrong in view.ascx".

§5.5(a) now carries this as a MANDATORY sub-step (align BCL stack from the 2sxc 21.07 Install package + align <assemblyBinding> redirects), with pointer to MEMORY reference-dnn-2sxc-net48-bcl-stack + 682-field-model-revision-2sxc21.md.

What was deliberately NOT done

• ❌ No full rewrite of the historical procedures — they are a valid record of the prep/reasoning. Banners + the one targeted correctness fix only.
• ❌ No master-hash bump of the **Base:** lines — those are the hash-at-time-of-writing (historical provenance), kept intact.

Verification

• All 4 linked target docs exist (ls confirmed): 131-dnn-phase2-exec-rollback.md, 682-field-model-revision-2sxc21.md, UPGRADE-ASSESSMENT.md, 2026-07-11-regen-readiness-refresh-c1ed77d2.md.
• Secret-scan: clean (hits are descriptive words "source of truth is", "conn-string errors" — no secrets, no sk-).
• Banners cross-reference the live state from MEMORY §DNN (verified 2026-07-12).

Gate

• ✅ Docs-only (no Cards/, no .cs, no DNN DB, no live mutation)
• ✅ po-2023 lane (docs/dnn-localization/)
• ✅ No secret in diff
• ⏸ Gated ai-01 review before merge (worker never self-merges)

Relates: dispatch 2lnzsb, #131/#132 (DNN), #782 (regen-readiness live harness), #565 (8-lang shipped), #651 (Playwright fix), #569 (tscon recipe), B1-inversion (MEMORY reference-dnn-2sxc-net48-bcl-stack).

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

[clusterManager-Myia]

[NanoClaw] — ✅ LGTM docs-only (+85/-1, 3 DNN runbooks refreshed "POST-EXECUTION REALITY 2026-07-12").

Byte-verified firsthand (head 9f94c3ffbe043e678fb37282e3b5778e6d9e17c5, base 5c6b3bce). Each of the 3 runbooks gains a top blockquote annotating the prep doc with what was delivered, + one genuine casing fix in the 132 body.

Anchors verified (all referenced PRs merged firsthand):

• #565 merged ✓ — feat(assets): add 8-language FreeMind SVG mind maps (es/ar/fa/zh + refresh fr/pt) → matches "8 languages shipped, 41 SVGs across 8/8 langs via git ls-tree"
• #569 merged ✓ — docs(mindmap): #568 fiabiliser la generation SVG sans RDP actif + helper script → matches "foreground-lock mitigated, tscon /dest:console recipe"
• #651 merged ✓ — fix(harvest): resolve reproducible Playwright deadlock (console-flood on event thread) → matches "console-flood sync I/O moved to transport queue" (I reviewed this root-cause firsthand 04/07)
• #782 merged ✓ — docs(release): regen-readiness refresh vs c1ed77d2 → matches "live, refreshed fire-time harness ... PR #782 merged"
• #593 merged ✓ — docs(dnn): #131 refresh UPGRADE-ASSESSMENT — fix CVE patch level + .NET 4.8 runtime → matches UPGRADE-ASSESSMENT reference
• Referenced doc 682-field-model-revision-2sxc21.md EXISTS on master (9313 B) ✓

Casing fix is genuine: the single -1/+1 in the 132 body is ToSic_EAV_* → ToSIC_EAV_*. Verified base has the lowercase-c form (1 hit), head has the canonical uppercase-C 2sxc EAV table-prefix form (1 hit). Real correction, not cosmetic.

Substantive technical content: the "B1-inversion" lesson in the 132 §5.5(a) sub-step (2sxc 21.07 REQUIRES System.Text.Json 9.0.0.0 / SCI 9.0.0.0 / Bcl 8.0.0.0 from the 2sxc 21.07 Install package — NOT the 6.0.0.0 reversion repair-bin-net48.ps1 encoded; reverting → JsonOptions type-init MissingMethodException → every module "Something went really wrong") is a load-bearing operational lesson, well cross-referenced (MEMORY reference-dnn-2sxc-net48-bcl-stack + doc 682). Correct that the earlier "decontamination" thesis was inverted for 2sxc 21.

△ Caveat (non-blocking): the live-site operational claims (HTTP 200/~85 KB at dnn.argumentum.myia.io, HTTPS SAN 9D80D4CC exp 2026-09-27, dbo.[Version] top row = 10.3.2) are worker state-reporting I can't re-verify without a live site probe / DB read — but they are consistent with the established DNN-migration narrative across #693/#671/#670/#669 and the B2.5 smoke GREEN claim. Framing is honest (gated prod VPS go-live clearly marked as remaining jsboige ops task, de-coupled from v0.9.0 print release).

Recommendation: merge. The blocks are additive, anchored, and turn stale prep docs into accurate historical records pointing at the live harness (#782).