diff --git a/.agents/skills/webjs/references/testing.md b/.agents/skills/webjs/references/testing.md
index a40bfff07..ec1b17507 100644
--- a/.agents/skills/webjs/references/testing.md
+++ b/.agents/skills/webjs/references/testing.md
@@ -36,7 +36,7 @@ Cancel a form on its `submit` event, not on the submit control's `click`. The fo
 
 Resolve the anchor from `e.composedPath()`, not `e.target.closest('a[href]')`. The listener is on `window`, so a click inside a shadow root (a `static shadow = true` component) arrives retargeted to the host, and `closest()` walks only light-tree ancestors and never finds the link, which fails open exactly where the router itself handles the case. A pure-fragment `href="#x"` link needs no guard at all, since it never navigates the page away.
 
-One thing this cannot cover: the router assigns `location.href` when it degrades a soft navigation, and `preventDefault` does not cancel a script assignment. Listen for `webjs:navigation-fallback` on `document` and assert none fired; its `cause` is the diagnosis.
+There is a SECOND channel a click guard cannot reach: the router assigns `location.href` when it degrades a soft navigation, and `preventDefault` does not cancel a script assignment. Do not try to intercept `location.href` itself, which is non-configurable on Chromium, Firefox, and WebKit alike, so its setter cannot be redefined on any of them. Listen for `webjs:navigation-fallback` on `document` and assert none fired; its `cause` is the diagnosis.
 
 It is a few lines, so write it in your suite's setup and detach it in teardown:
 
@@ -51,7 +51,7 @@ window.addEventListener('click', onClick);     // bubble, never capture
 window.addEventListener('submit', onSubmit);
 ```
 
-(The WebJs framework repo keeps its own copy of exactly this in one shared module, `test/browser-nav-guard.js`, whose `installNavGuard()` returns `{ fallbacks, remove }`. That module is framework-repo infrastructure and is not part of a scaffolded app.)
+(The WebJs framework repo keeps its own copy of exactly this in one shared module, `test/browser-nav-guard.js`, whose `installNavGuard()` returns `{ fallbacks, hardNavigations, remove }`. That module is framework-repo infrastructure and is not part of a scaffolded app.)
 
 ## App runners (`webjs test`)
 
diff --git a/packages/core/AGENTS.md b/packages/core/AGENTS.md
index f4f3d2ed5..2e95589ec 100644
--- a/packages/core/AGENTS.md
+++ b/packages/core/AGENTS.md
@@ -238,7 +238,7 @@ Chromium, Firefox, and WebKit via web-test-runner).
 
 A browser test that clicks a real `<a href>` or submits a real `<form>`
 MUST install the shared guard from `test/browser-nav-guard.js`
-(`installNavGuard()`, returning `{ fallbacks, remove }`). web-test-runner
+(`installNavGuard()`, returning `{ fallbacks, hardNavigations, remove }`). web-test-runner
 aborts the whole session when the page navigates, so an escaped click
 takes down every browser file rather than failing one test. The guard
 listens on `window` in the BUBBLE phase, which is load-bearing: capture
@@ -247,6 +247,17 @@ runs, and `onClick` returns on that flag, so every guarded router test
 would pass while testing nothing. Full rule in
 [`references/testing.md`](../../.agents/skills/webjs/references/testing.md).
 
+The guard also covers the SECOND channel, the router's own hard navigation on
+a degradation. `preventDefault` cannot cancel a script assignment, and
+`location.href` is non-configurable on all three engines so its setter cannot
+be redefined, so `router-client.js` routes every hard navigation through one
+`hardNavigate` indirection with a `setHardNavigate(fn)` seam. The guard
+installs an override that records the attempt into `hardNavigations` instead of
+performing it, so a degradation fails one test with its `cause` slug rather
+than aborting the whole session. `setHardNavigate` is TEST-ONLY and is
+deliberately not re-exported from `index.js` / `index-browser.js`; tests reach
+it through the same direct `src/router-client.js` import they already use.
+
 Cross-package tests that exercise core through the SSR pipeline
 or scaffolds live at the repo root in `test/ssr/`,
 `test/scaffolds/`, etc. See [`references/testing.md`](../../.agents/skills/webjs/references/testing.md).
diff --git a/packages/core/src/router-client.js b/packages/core/src/router-client.js
index 10bcfac0a..54933fb89 100644
--- a/packages/core/src/router-client.js
+++ b/packages/core/src/router-client.js
@@ -377,7 +377,9 @@ export function disableClientRouter() {
 export async function navigate(url, opts) {
   const target = new URL(url, location.href);
   if (target.origin !== location.origin) {
-    location.href = url;
+    // Cross-origin: an intentional full-page nav, not a degradation, but it
+    // ends the session in a test just the same, so it rides the same seam.
+    hardNavigate(url);
     return;
   }
   await performNavigation(target.href, opts?.replace ?? false, null);
@@ -475,6 +477,28 @@ export function revalidate(url) {
  */
 const NON_HTML_EXTENSIONS = /\.(?:pdf|zip|tar|gz|7z|rar|dmg|exe|msi|deb|rpm|apk|ipa|xlsx?|docx?|pptx?|csv|odt|ods|odp|rtf|epub|mobi|xml|json|rss|atom|txt|md|wasm|mp3|mp4|mov|avi|webm|ogg|flac|wav|m4a|m4v|mkv|png|jpe?g|gif|webp|avif|bmp|ico|svg|tiff?|heic)$/i;
 
+/**
+ * The one place the router hands a navigation back to the browser.
+ *
+ * Every hard navigation the router performs goes through here rather than
+ * assigning `location.href` inline, so a browser test can observe it. The
+ * default is exactly the assignment it replaces, so behaviour is unchanged
+ * unless something calls `setHardNavigate`.
+ *
+ * This exists because a hard navigation is UNOBSERVABLE and UNPREVENTABLE from
+ * outside. `preventDefault` cancels a default action, not a script assignment,
+ * and `location.href` is non-configurable on Chromium, Firefox, and WebKit
+ * alike, so a test cannot redefine its setter either (measured; the older
+ * `spyOnReload` helper that tried was silently a no-op on every engine). In a
+ * web-test-runner session a real navigation aborts the WHOLE session, so one
+ * degradation destroys every remaining browser test file and reports `0 failed`
+ * on the way out. A seam is the only thing that makes it catchable.
+ *
+ * @param {string} href
+ */
+let hardNavigate = (href) => { location.href = href; };
+
+
 /** @param {MouseEvent} e */
 function onClick(e) {
   if (!enabled) return;
@@ -1181,7 +1205,7 @@ async function performNavigation(href, isPopState, frameId) {
   // nav carries its own boundary element.
   if (shouldFullLoadDuringParse(isPopState, frameId) && typeof location !== 'undefined') {
     reportFallback('readyState-loading', href);
-    location.href = href;
+    hardNavigate(href);
     return;
   }
 
@@ -2082,7 +2106,7 @@ function handleNavigationError(href, status, error) {
   // carries no `cause` / `willReload`, so it is not a substitute.
   if (typeof location !== 'undefined') {
     reportFallback('navigation-error-unrecoverable', href);
-    location.href = href;
+    hardNavigate(href);
   }
 }
 
@@ -2910,14 +2934,14 @@ function applySwap(doc, frameId, revalidating, href, incomingBuild, incomingSrc)
         } else {
           if (sessionStorage) sessionStorage.setItem(flag, '1');
           reportFallback('deploy-mismatch', href);
-          location.href = href;
+          hardNavigate(href);
           return;
         }
       } catch {
         // sessionStorage unavailable (private mode w/ quota etc.):
         // fall through to a single reload like before.
         reportFallback('deploy-mismatch', href);
-        location.href = href;
+        hardNavigate(href);
         return;
       }
     } else if (!mismatch) {
@@ -3055,7 +3079,7 @@ function applySwap(doc, frameId, revalidating, href, incomingBuild, incomingSrc)
     reportFallback(!here ? 'live-boundaries-malformed'
       : !there ? 'incoming-boundaries-malformed'
       : 'no-shared-boundary', href);
-    location.href = href;
+    hardNavigate(href);
     return;
   }
 
@@ -4523,6 +4547,20 @@ export function _resetPrefetch() {
   clearPrefetchViewTimers();
 }
 
+/**
+ * Test-only: replace the hard-navigate action so a browser test can observe a
+ * navigation instead of being destroyed by it. Call with no argument to
+ * restore. Underscore-prefixed and kept in this block like every other
+ * test-only export here, so it stays out of `router-client.d.ts` and out of
+ * the app-facing API (the `./client-router` subpath resolves this file under
+ * the `source` condition, so an unprefixed name here would read as public).
+ *
+ * @param {((href: string) => void) | null} [fn]
+ */
+export function _setHardNavigate(fn) {
+  hardNavigate = fn || ((href) => { location.href = href; });
+}
+
 /** Test-only: read the monotonic navigation-token counter. */
 export function _navToken() { return currentNavigationToken; }
 /** Test-only: bump the navigation-token counter (simulates a fresh nav). */
diff --git a/packages/core/test/routing/browser/form-action-submit.test.js b/packages/core/test/routing/browser/form-action-submit.test.js
index 1b47628be..70e8e7646 100644
--- a/packages/core/test/routing/browser/form-action-submit.test.js
+++ b/packages/core/test/routing/browser/form-action-submit.test.js
@@ -36,10 +36,6 @@ suite('Client router: bound form submissions (#1155)', () => {
   let navGuard;
 
   let container, origFetch, calls;
-  // When a test redefines window.location.href (to detect a full-page reload),
-  // it records the restore fn here so teardown reverts it even if the body
-  // throws. Null when no redefine is active.
-  let restoreHref;
 
   let bOpen, bClose;
   function setup(responder) {
@@ -55,7 +51,6 @@ suite('Client router: bound form submissions (#1155)', () => {
     document.body.appendChild(container);
     document.body.appendChild(bClose);
     calls = [];
-    restoreHref = null;
     origFetch = window.fetch;
     window.fetch = (url, init) => {
       calls.push({ url: String(url), init: init || {} });
@@ -65,38 +60,11 @@ suite('Client router: bound form submissions (#1155)', () => {
   function teardown() {
     navGuard.remove();
     window.fetch = origFetch;
-    if (restoreHref) { try { restoreHref(); } catch { /* ignore */ } restoreHref = null; }
     container.remove();
     if (bOpen) bOpen.remove();
     if (bClose) bClose.remove();
   }
 
-  /**
-   * Replace window.location.href's setter with a spy so a full-page reload is
-   * observable (the router falls back to `location.href = url` only for a
-   * non-HTML / error response). Returns a getter for the reload count. The
-   * descriptor restore is registered on `restoreHref` so teardown always
-   * reverts it. Some browsers forbid redefining the accessor; in that case the
-   * spy is a no-op and the test leans on the DOM-applied assertion instead.
-   */
-  function spyOnReload() {
-    let reloads = 0;
-    const realDescriptor = Object.getOwnPropertyDescriptor(Location.prototype, 'href')
-      || Object.getOwnPropertyDescriptor(window.location, 'href');
-    let installed = false;
-    try {
-      Object.defineProperty(window.location, 'href', {
-        configurable: true,
-        get: () => location.toString(),
-        set: () => { reloads += 1; },
-      });
-      installed = true;
-    } catch { /* redefining forbidden here; rely on the DOM assertion */ }
-    if (installed && realDescriptor) {
-      restoreHref = () => Object.defineProperty(window.location, 'href', realDescriptor);
-    }
-    return { count: () => reloads, installed: () => installed };
-  }
 
   test('a bound form posts to the page own url and carries the identity field', async () => {
     // The rendered form has NO `action` attribute (the renderer omits it so the
@@ -161,7 +129,6 @@ suite('Client router: bound form submissions (#1155)', () => {
       '<input name="email" value="bad"></form></main><!--/wj:children:/-->',
       { status: 422, headers: { 'content-type': 'text/html', 'x-webjs-build': '' } },
     ));
-    const reload = spyOnReload();
     try {
       render(html`
         <main>
@@ -176,7 +143,12 @@ suite('Client router: bound form submissions (#1155)', () => {
       await tick();
 
       assert.ok(calls.length, 'fetch was issued');
-      assert.equal(reload.count(), 0, '422 HTML must be applied in place, never a full reload');
+      // The seam (#1286) records a hard navigation instead of performing it, so
+      // this is a real observation. The old `spyOnReload` helper could not make
+      // it: `location.href` is non-configurable on all three engines, so its
+      // redefine always threw and the count was structurally always zero.
+      assert.equal(navGuard.hardNavigations.length, 0,
+        '422 HTML must be applied in place, never a full reload');
       // The 422 body was actually applied to the live DOM (the field error is
       // now present), which a full reload would never achieve from a fetch stub.
       assert.ok(document.getElementById(marker), 'the 422 re-render body was applied in place');
diff --git a/packages/core/test/routing/browser/nav-guard.test.js b/packages/core/test/routing/browser/nav-guard.test.js
index c3a72c216..fa5daf4ef 100644
--- a/packages/core/test/routing/browser/nav-guard.test.js
+++ b/packages/core/test/routing/browser/nav-guard.test.js
@@ -137,6 +137,45 @@ suite('Browser-test nav guard (#1135)', () => {
     } finally { teardown(); }
   });
 
+  test('a real degradation is recorded, not performed (#1286)', async () => {
+    setup();
+    try {
+      // Force the router to degrade: strip the live boundary pair so the swap
+      // cannot find a shared boundary. That path reports a fallback and then
+      // hands the navigation to the browser, which before the seam existed
+      // aborted the whole web-test-runner session rather than failing here.
+      bOpen.remove();
+      bClose.remove();
+      render(html`<a href="/nav-guard-degrade">go</a>`, container);
+      const settled = awaitNavigation();
+      container.querySelector('a').click();
+      await settled;
+
+      assert.ok(guard.hardNavigations.some((u) => u.includes('/nav-guard-degrade')),
+        'the hard navigation must be RECORDED by the seam');
+      // There is deliberately NO in-test assertion that the navigation was not
+      // PERFORMED, because none can be honest. `location` cannot serve: the
+      // degradation path falls through to `history.pushState`, so the pathname
+      // changes either way. Nor can a surviving window sentinel: a
+      // `location.href` assignment starts a navigation that commits on a later
+      // task rather than tearing the realm down synchronously, so the sentinel
+      // reads the same on both sides and would be a vacuous assertion, which is
+      // the exact defect class this seam exists to remove.
+      //
+      // What proves non-performance is the counterfactual: disable the override
+      // in `installNavGuard` and this file does not fail, it aborts the whole
+      // web-test-runner session on every engine.
+      // The cause slug is the diagnosis, and it only survives because the
+      // navigation no longer happens.
+      assert.ok(guard.fallbacks.length > 0, 'the degradation reported a cause');
+      assert.match(String(guard.fallbacks[0].cause), /boundar|shared/,
+        `expected a boundary-related cause, got ${guard.fallbacks[0].cause}`);
+    } finally {
+      // teardown() removes bOpen/bClose; they are already detached here.
+      teardown();
+    }
+  });
+
   test('does NOT suppress the router on a plain link (capture-phase regression)', async () => {
     setup();
     try {
diff --git a/test/browser-nav-guard.js b/test/browser-nav-guard.js
index 85ecba8f1..8e5a06aa9 100644
--- a/test/browser-nav-guard.js
+++ b/test/browser-nav-guard.js
@@ -1,3 +1,5 @@
+import { _setHardNavigate } from '../packages/core/src/router-client.js';
+
 /**
  * Shared navigation guard for browser tests (#1135). Sibling of
  * `test/browser-assert.js` (#777) and with the same "one source of truth for
@@ -37,21 +39,40 @@
  * possible failure here, and it is why `nav-guard.test.js` asserts the router
  * still ran rather than trusting this by inspection.
  *
- * ## What it cannot do
+ * ## The second channel: the router's own hard navigation
+ *
+ * `preventDefault` cancels a default action, not a script assignment, so it can
+ * never stop the router assigning `location.href` when it degrades. That is a
+ * separate channel and it needs a separate mechanism: the router routes every
+ * hard navigation through one `_setHardNavigate` seam (#1286), and this installs an override that
+ * RECORDS the attempt into `hardNavigations` instead of performing it. So a
+ * degradation now fails the one test with a readable message instead of
+ * aborting the whole session.
+ *
+ * Intercepting `location.href` directly is not an option and should not be
+ * attempted: it is non-configurable on all three engines, so its setter cannot
+ * be redefined. That is why the seam lives in the router rather than here.
  *
- * It cannot stop the router's own `location.href` assignment on a degradation:
- * `preventDefault` cancels a default action, not a script assignment. The
- * `fallbacks` array is the coverage for that second channel. Every reload site
- * dispatches `webjs:navigation-fallback` with a stable `cause` immediately
- * beforehand, so a test asserts `fallbacks` is empty and names the cause.
- * Removing the conditions that cause a degradation is the fixture work in
- * #1053.
+ * `fallbacks` stays useful alongside it: it carries the stable `cause` slug
+ * that says WHY the router degraded, which the recorded href alone does not.
  *
- * @returns {{ fallbacks: Array<{cause: string, href: string, willReload: boolean}>, remove: () => void }}
+ * This catches one navigation that is NOT a degradation: a cross-origin
+ * `navigate()`, which is an intentional full-page nav. It is recorded rather
+ * than performed like any other, so it is observable (assert on
+ * `hardNavigations`) instead of ending the session. Nothing is swallowed
+ * silently; a suite asserting `hardNavigations` is empty will fail on it.
+ *
+ * Note this module imports the router, which self-enables on load. Every suite
+ * that installs the guard is already a router suite that imports it, so this
+ * changes nothing in practice.
+ *
+ * @returns {{ fallbacks: Array<{cause: string, href: string, willReload: boolean}>, hardNavigations: string[], remove: () => void }}
  */
 export function installNavGuard() {
   /** @type {Array<{cause: string, href: string, willReload: boolean}>} */
   const fallbacks = [];
+  /** @type {string[]} */
+  const hardNavigations = [];
 
   const onClick = (e) => {
     // Walk the COMPOSED path, exactly as the router's `findAnchorInPath` does,
@@ -76,13 +97,18 @@ export function installNavGuard() {
 
   const onFallback = (e) => { fallbacks.push(e.detail); };
 
+  // Record the router's own hard navigations instead of performing them.
+  _setHardNavigate((href) => { hardNavigations.push(String(href)); });
+
   window.addEventListener('click', onClick);
   window.addEventListener('submit', onSubmit);
   document.addEventListener('webjs:navigation-fallback', onFallback);
 
   return {
     fallbacks,
+    hardNavigations,
     remove() {
+      _setHardNavigate(null);
       window.removeEventListener('click', onClick);
       window.removeEventListener('submit', onSubmit);
       document.removeEventListener('webjs:navigation-fallback', onFallback);
diff --git a/website/app/docs/testing/page.ts b/website/app/docs/testing/page.ts
index 98d36e9d2..12aa4c014 100644
--- a/website/app/docs/testing/page.ts
+++ b/website/app/docs/testing/page.ts
@@ -231,7 +231,7 @@ const onSubmit = (e) =&gt; e.preventDefault();   // forms cancel on submit, not
 window.addEventListener('click', onClick);    // window BUBBLE phase, never capture
 window.addEventListener('submit', onSubmit);</code-block>
     <p>Register on <code>window</code> in the <strong>bubble</strong> phase. That is the last step of the propagation path, so it runs after the router's own listeners, and <code>preventDefault()</code> still cancels the default action. Never use the capture phase: it sets <code>defaultPrevented</code> before the router sees the event, and the router bows out on that flag, so every such test would pass while testing nothing. A pure-fragment <code>href="#x"</code> link needs no guard, since it never navigates the page away.</p>
-    <p>One case this cannot cover: the router assigns <code>location.href</code> when it degrades a soft navigation, and <code>preventDefault</code> does not cancel a script assignment. Listen for <code>webjs:navigation-fallback</code> on <code>document</code> and assert none fired; its <code>cause</code> names the reason.</p>
+    <p>There is a second channel a click guard cannot reach: the router assigns <code>location.href</code> when it degrades a soft navigation, and <code>preventDefault</code> does not cancel a script assignment. Do not try to intercept <code>location.href</code> itself, which is non-configurable on Chromium, Firefox, and WebKit alike, so its setter cannot be redefined on any of them. Listen for <code>webjs:navigation-fallback</code> on <code>document</code> and assert none fired; its <code>cause</code> names the reason.</p>
 
     <h2>Convention Validation</h2>
     <p><code>webjs check</code> validates your app for correctness issues:</p>