design-b D0: record the audit + red pin in docs 19/20/21

Doc 19: status DIAGNOSED -> DIAGNOSED + PINNED RED; the repro section
leads with the automated deterministic spec (the manual Leonardo flow
stays as the field repro).
Doc 20: D0 work-log entry; status D-1 + D0 done; next = D1.
Doc 21: red-spec section updated to "landed", with the staging test's
anti-vacuity role.

Carried finding: the strand reproduces on a 2-OBJECT fixture schematic,
so warm-load byte volume (the 68/1 dice-loader) is not an ingredient —
two concurrent parks suffice. That is why D3 closes this class and D5 is
not required for it.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01TEHGiiXMShNXbBr7gSJ7iz

docs/features/async/19-quasimodal-fiber-strand.md

@@ -1,14 +1,27 @@
 # 19 — Symbol Properties dialog hangs: a stranded tool fiber (live investigation)
 
-> **Status: DIAGNOSED, NOT FIXED (2026-08-05).** Reproduced end-to-end in a real browser
-> against the dev platform on the scheduler build. Regression-vs-pre-existing is **not yet
+> **Status: DIAGNOSED + PINNED RED, NOT FIXED (2026-08-05).** Reproduced end-to-end in a
+> real browser against the dev platform on the scheduler build, and since D0 also
+> **deterministically in e2e**: `tests/kicad/quasimodal-strand.spec.ts` (6/6 identical —
+> `closed=false dialogs=1 refused-resumes=1`). The fix lands at doc 20 **D3**, where that
+> spec's `test.fail()` marker comes off. Regression-vs-pre-existing is **not yet
 > determined** — see §5. Related: [`17`](17-mailbox-scheduler-plan.md) S4 (waits),
 > [`16`](16-fiber-resume-guard.md) (the quarantine guard), the 8/4 three-UI-bugs triage
 > (which blamed the interlock drain — that is the *symptom layer*, not the proximate cause).
 
-## 1. Repro (100%, ~40 s)
+## 1. Repro
 
-Dev platform (`npm run dev`), editor at `:3048`, Arduino Leonardo schematic:
+**Automated (D0, ~30 s, deterministic):**
+`npx playwright test --project=kicad-firefox kicad/quasimodal-strand.spec.ts`. The
+"staging" test proves the window is real (dialog open + timer fired + concurrent-park
+beacons); the "doc-19 red" test clicks OK and records the hang. The overlap is
+structural, not a race: the parking timer (`wasm/bindings/timer_park.h`) is armed
+*after* the dialog is confirmed open, so it necessarily parks on top of the opener's
+open-ended fiber park. Note the strand reproduces on a **2-object fixture schematic** —
+Leonardo's byte volume is not an ingredient, only concurrent parks are.
+
+**Manual (100%, ~40 s):** dev platform (`npm run dev`), editor at `:3048`, Arduino
+Leonardo schematic:
 
 1. Open the project URL, wait for load (~30 s).
 2. Double-click the USB receptacle (J1) → **Symbol Properties** opens.

docs/features/async/20-design-b-core-plan.md

@@ -1,6 +1,6 @@
 # 20 — Design B core: parkable activities as scheduler contexts
 
-> **Status: IN PROGRESS — D-1 DONE (2026-08-05, work log §10).** The remaining core of
+> **Status: IN PROGRESS — D-1 + D0 DONE (2026-08-05, work log §10).** The remaining core of
 > Design B ([`06`](06-design-b-fiber-first-runtime.md) B1+B2), scoped against what S0–S6
 > actually built ([`17`](17-mailbox-scheduler-plan.md)) and motivated by the
 > stranded-fiber hang ([`19`](19-quasimodal-fiber-strand.md)). Supersedes doc 12's
@@ -218,4 +218,29 @@ occ-probe `glb` case (known-unrelated, predates D-1). Vestigial `startModal` scr
 from `ASYNCIFY_IMPORTS` (Makefile.wasm) and `asyncify-imports.txt` post-gate.
 
 Consequence: every later phase is single-path — one code path per park site, one glue
-per build, one battery per gate. Next: **D0** (park-site audit + the doc-19 red spec).
+per build, one battery per gate.
+
+### D0 — park-site audit + red spec (2026-08-05) ✅
+
+- **Audit** → [`21`](21-park-site-audit.md) (pcbjam `d4b25a0`): 14 production park sites
+  (8 wx, 9 KiCad/bridge) + 3 deliberate test levers, each classified fiber /
+  entry-stack / main-loop with its routing phase. W1 `wxWasmYieldUntilJs` and W3 popup →
+  D3; the clipboard/font/lib/3D/occ/ngspice/nanosleep bridges → D4; W2's per-frame yield
+  stays safe-by-construction unless D5 is taken. Doc 20 risk 4 (pthreads) settled: all
+  Asyncify parks are main-thread-only; the lib bridge's worker path is a blocking
+  proxy, not a park, so contexts stay a main-thread concept.
+- **Red spec** → `tests/kicad/quasimodal-strand.spec.ts` (pcbjam `6ab0843`): a GREEN
+  "staging" test (the window is real) plus the `test.fail()` doc-19 pin (OK must close
+  the dialog). 6/6 identical: `closed=false dialogs=1 refused-resumes=1`. The overlap is
+  structural — the parking timer is armed *after* the dialog opens, on top of the
+  opener's open-ended fiber park.
+- **Finding worth carrying into D1:** the strand reproduces on a **2-object fixture
+  schematic**. Byte volume (the doc-17/gal-refresh dice-loader) is NOT an ingredient
+  here; two concurrent parks suffice. That makes the doc-19 class strictly easier to
+  hit than the 68/1 warm-load family, and it is why D3 — not D5 — is the phase that
+  closes it.
+- Retired `tests/kicad/dialog-deadlock-probe.spec.ts` (the 8/4 throwaway probe, and the
+  tree's only determinism-lint violations).
+
+Next: **D1** — context primitives on libcontext, exercised by a dedicated test app, no
+production path switched, with the context-count/peak-RSS gate doc 20 §7 risk 1 asks for.

docs/features/async/21-park-site-audit.md

@@ -87,9 +87,17 @@ belongs in the scheduler shim (it already owns `Fibers` bookkeeping and the
 
 ## 4. Red spec (the other D0 deliverable)
 
-The doc-19 strand lands as `tests/kicad/quasimodal-strand.spec.ts`: open a schematic,
-double-click a symbol → Symbol Properties (quasi-modal on a tool fiber = W1's fiber
-lane), click OK → the dialog must close and dispatch must stay live. RED today by the
-doc-19 mechanism; goes green at D3. Marked `test.fail()` so the battery stays runnable
-while red — when D3 lands, Playwright flags "expected to fail but passed", forcing the
-flip to a plain green pin.
+**Landed** as `tests/kicad/quasimodal-strand.spec.ts` (pcbjam `6ab0843`): open a
+schematic, double-click a symbol → Symbol Properties (quasi-modal on a tool fiber =
+W1's fiber lane), arm the parking timer *while the dialog is up* (so its park lands on
+top of the opener's open-ended park — structural overlap, not a race), click OK → the
+dialog must close and the wait books must balance. RED today by the doc-19 mechanism,
+6/6 identical (`closed=false dialogs=1 refused-resumes=1`); goes green at D3. Marked
+`test.fail()` so the battery stays runnable while red — when D3 lands, Playwright flags
+"expected to fail but passed", forcing the flip to a plain green pin. A separate GREEN
+"staging" test asserts the window is real, so the pin cannot rot into vacuity.
+
+**Audit consequence recorded there:** the strand reproduces on a 2-object fixture — the
+warm-load byte volume that dice-loads the 68/1 family is not an ingredient. Two
+concurrent parks suffice, which is why D3 (waits on contexts) closes this class and D5
+(main as a context) is not required for it.