fix(asks): one submit saves every ask on the page

Operator report (via infra-ops): on a page with several asks, a submit
saved only the pressed one and the reload wiped the rest. Confirmed on
auk-audition: one POST at 15:02:23 saved the last ask on the page, then a
400 from the submit of an ask the reload had just blanked.

Client-side on both surfaces; /answer is unchanged. A submit on a pick
form, while another pick form on the page holds unsent input, sends every
changed ("dirty") pick form: one POST each, to its own action, with
Accept: application/json, in document order. A refusal stops nothing, and
untouched forms are never re-sent. With no other dirty form, a submit is
exactly what it was.

- embed.js (verbatim reports): reloads only when nothing was refused and
  nothing of ours is dirty. Otherwise a server-rendered status line in the
  submit block says what did not save, and input stays. A form the server
  took gets a new baseline. A press during the flight is ignored.
- base.html (marks page, lightbox, review rail): one refresh in place. A
  batch never reloads. Only forms the server took count as sent. In-flight
  state and "just sent" are keyed by form identity (formKey) plus the fields
  at the press, not the DOM node.

Two heid bug-hunt rounds: a four-arm panel on the first cut, then Hulda
alone on the fold. Ten findings reproduced red in a browser before their
fixes. Contracts: U3 "Submitting several asks at once" + INV-8, R2 C3
steps 2, 3 and 3a. Mutation tables u3_submit_all (15) and r2_submit_all
(11), all proved. Suite 928 -> 951.

persistent-memory.md

@@ -1,6 +1,6 @@
 # Persistent memory — booth
 
-_Last updated: 2026-09-25_
+_Last updated: 2026-09-27_
 
 > **Always check for `/tmp/booth-dev-handoff.md`** — if it exists and its
 > `Written:` stamp is under 8 hours old, read it (it carries the in-flight
@@ -17,12 +17,14 @@ loop it turned out to actually be.
 
 ## Current state / in-flight
 
-_As of 2026-09-25:_
+_As of 2026-09-27:_
 
-- 🟢 **NOTHING IS IN FLIGHT** (shutdown snapshot, 2026-09-25). The tree is
-  clean, no branch or worktree is open, and no peer is waiting on booth-dev.
-  main is ONE memory commit ahead of origin (this snapshot), unpushed: the
-  push is the operator's call.
+- ✅ **ONE SUBMIT SAVES EVERY ASK ON A PAGE** (2026-09-27, Prime's bug via
+  infra-ops, thread `01M3JED397G1SZH7580PCXNVVA`). Client-side on both
+  surfaces (embed.js, base.html's in-place script), no server change; a
+  refusal or a mid-flight edit never clears input. Local commits on main,
+  UNPUSHED (push is the operator's call); service restarted to make it live.
+  → `persistent-memory.d/2026-09-27-one-submit-saves-every-ask.md`
 - ✅ **r3 COMPARE IS LIVE AND PUSHED** (operator: "merge r3 once the mutation
   check is clean", then "merge and push", 2026-09-24). origin/main is the r3
   arc's tip: r3 (`f8d136a`), design-dev's race fix (`d54bb04`: one compare ring
@@ -85,6 +87,7 @@ _As of 2026-09-25:_
 
 ## Recent decisions
 
+- `[2026-09-27]` ✅ **One submit saves every ask on the page; the heid panel found the async window** — READ BEFORE TOUCHING THE SUBMIT PATH OF embed.js OR base.html: a batch reads forms at the press while the page stays live, and only a test that acts inside the flight can see it → `persistent-memory.d/2026-09-27-one-submit-saves-every-ask.md`
 - `[2026-09-24]` ⏸ **The upload route's three lifecycle gaps: DEFERRED** — the pickup-id `mkdir` sits outside the try (a FileExistsError race), `rmtree(ignore_errors=True)` hides its own failure, and `except Exception` misses CancelledError. All three are rare; the operator was told and merged without them. Tracked in `225ba32`'s commit message.
 - `[2026-09-24]` ✅ **Upload names: two crashes found, then two holes in the fix** — READ BEFORE WRITING A SANITISER: drop everything droppable FIRST, then apply the structural rules; a NUL test through httpx `files=` proves nothing → `persistent-memory.d/2026-09-24-upload-names-two-crashes-then-two-holes.md`
 - `[2026-09-24]` ✅ **The r3 seam pass: what only it could see** — 12 contract-vs-code mismatches folded before code. READ BEFORE A CONTRACT THAT ADDS `data-region`S, MOVES TEMPLATE CODE, OR ADDS A `/b/{name}/<word>` ROUTE → `persistent-memory.d/2026-09-24-r3-seam-pass-what-only-it-could-see.md`