fix(harness): make a vacuous watchdog denial fail, and single-home the window

PR #65's review found this PR had planted a fresh instance of the defect it
exists to repair. Both blocking findings fixed, each demonstrated.

FINDING 1. "a peer that cannot be reached is named, and the watchdog stays out
of it" could not see the property its comment claimed. `shell.arm` runs only
after `await transport.ready`, which never resolves on a socket that never
opens — so the watchdog is never armed there and denying "sent no document"
denies an unreachable event. Reverting PR #21's routing left that clause
passing.

The aggravating half: `context/test-evidence/clients.md` already said so, and
this PR EDITED that paragraph — inserting "runFor(10s) advances the watchdog"
immediately above the sentence explaining it is never armed. Not text the PR
left alone; text the PR walked past mid-sentence.

The clause is gone and the test keeps the naming property it really guards: a
`close` on a socket that never opened must not read as "disconnected". And the
shape cannot return — `outlastTheWatchdog` now asks the page `timesArmed() > 0`
before jumping, so a watchdog denial re-added to that path FAILS BY NAME. It
asks *ever armed* rather than *armed now* on purpose: by the time the disconnect
test jumps, a correct page has already disarmed, which is the property, so the
obvious spelling would have denied the passing case. Building the guard is what
surfaced that distinction; reading had not.

FINDING 2. The spec kept its own `HANDSHAKE_TIMEOUT_MS` beside a comment saying
a silent disagreement "would pass for the wrong reason". It did: the spec at
1000 against the page's 8000 made the suite green over the reverted routing,
because the jump landed inside the window. The constant is now exported from
`web/shell.js` and imported by both, so the disagreement is unconstructible.
Shrinking it in its one home still catches the routing revert — that is the
demonstration the guard did not go quiet.

Filed, not fixed: improvise-78v6 (#61 also deleted the help-overlay spec and
nothing replaced it — a PR about a dropped guard that did not diff the dropping
PR's test names) and improvise-5ray (--check-evidence collects only `### ` and
resolves only Rust fns, so every browser entry's `####` is invisible; appending
one that names nothing exits 0, which is how the dangling heading cleared the
gate for a release).

Verified on the merged tree: node 26/26, bundle 3/3, Playwright 17 passed + 3
skipped, cargo test --workspace 1378 passed, comment-runs clean,
--check-evidence 501 resolving, all cited ids resolve, no .beads diff.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01FC2jVn8Au7E9JPkw6bwXXH

web/app.js

@@ -4,6 +4,7 @@
 import init, { Client } from "./pkg/improvise_wasm_client.js";
 import {
   COL_PX,
+  HANDSHAKE_TIMEOUT_MS,
   ROW_PX,
   createShell,
   extentOf,
@@ -40,13 +41,15 @@ const client = new Client();
 
 // A peer that never answers the handshake, and the two facts `#status` shows.
 // Both are decided in shell.js; this is the wire to the element.
-const HANDSHAKE_TIMEOUT_MS = 8000;
 const shell = createShell({
   onStatus: (text) => {
     el.status.textContent = text;
   },
   timeoutMs: HANDSHAKE_TIMEOUT_MS,
 });
+// The harness asks whether the watchdog is ARMED, because a negative assertion
+// about it is vacuous where it is not (clients.md). Read by tests, nothing else.
+window.__improviseShell = shell;
 
 // What the grid's own frame costs, in whole cells, as of the last repaint —
 // measured off the painted model, never predicted. context/code/clients.md.

web/shell.js

@@ -27,6 +27,11 @@ export function indicator(collapsed) {
 export const ROW_PX = 26;
 export const COL_PX = 110;
 
+/** How long a peer has to answer the handshake. Exported because the harness
+ *  must jump *past* it: a spec with its own copy can understate it silently
+ *  and every watchdog assertion then passes vacuously (clients.md). */
+export const HANDSHAKE_TIMEOUT_MS = 8000;
+
 /** The extent a box of this size can show, in whole DATA cells: the frame is
  *  charged for first, and at least one cell is always asked for. */
 export function extentOf(box, frameCost, metrics = { rowPx: ROW_PX, colPx: COL_PX }) {
@@ -107,6 +112,7 @@ export function createShell({ onStatus, timeoutMs, setTimer = setTimeout, clearT
   let shell = "";
   let model = "";
   let watchdog = null;
+  let arms = 0;
   const paint = () => onStatus(composeStatus(shell, model));
 
   return {
@@ -121,6 +127,7 @@ export function createShell({ onStatus, timeoutMs, setTimer = setTimeout, clearT
     /** A peer that never answers the handshake. Deliberately about the *first
      *  frame*, not either peer's own failure modes — clients.md. */
     arm(message) {
+      arms += 1;
       clearTimer(watchdog);
       watchdog = setTimer(() => {
         watchdog = null;
@@ -140,5 +147,8 @@ export function createShell({ onStatus, timeoutMs, setTimer = setTimeout, clearT
       this.setShell(text);
     },
     armed: () => watchdog !== null,
+    /** How many times the watchdog has been armed. A test denying its message
+     *  needs this to be nonzero, or it is denying an unreachable event. */
+    timesArmed: () => arms,
   };
 }