diff --git a/.github/workflows/webkit-stall-experiment.yml b/.github/workflows/webkit-stall-experiment.yml
new file mode 100644
index 0000000000..7c0939efc2
--- /dev/null
+++ b/.github/workflows/webkit-stall-experiment.yml
@@ -0,0 +1,67 @@
+# TEMPORARY: experiment for a draft PR, never to be merged.
+# Does a WebKit page on Linux stall after rendering a million rows, or under the suite's load,
+# and do the stalls line up with drops in the memory of WebKit's web processes?
+name: WebKit stall experiment
+
+on: pull_request
+
+permissions: {}
+
+jobs:
+  experiment:
+    runs-on: ubuntu-latest
+    timeout-minutes: 30
+
+    permissions:
+      contents: read # check out the repository
+
+    strategy:
+      fail-fast: false
+      matrix:
+        run: [1, 2, 3]
+
+    steps:
+      - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7.0.1
+        with:
+          persist-credentials: false
+
+      - uses: voidzero-dev/setup-vp@68f39c78aaf39be468cabd2afb09c162574c84af # v1.21.2
+        with:
+          cache: true
+          cache-save: false
+          run-install: false
+
+      - run: vp install --frozen-lockfile
+
+      - run: vp run install-browsers
+        timeout-minutes: 8
+
+      - name: Idle after rendering a million rows, then 100k rows
+        env:
+          WEBKIT_STALL_SAMPLER: 1
+          WEBKIT_STALL_IDLE: 1
+          DEBUG: pw:browser*
+        run: |
+          (while sleep 0.2; do echo "$(date +%s%3N) $(ps -eo pid=,rss=,comm= | grep -E 'WebProces' | tr -s ' ' | tr '\n' ';')"; done) > memory-idle.log &
+          poller=$!
+          vp test test/browser/zzStall.experiment.tsx --project "browser (webkit)" --sequence.shuffle.tests=false > idle.log 2>&1 || true
+          kill "$poller"
+          grep -E 'STALLS|Tests |Duration' idle.log | cut -c1-2000
+
+      - name: WebKit suite with the stall sampler
+        env:
+          WEBKIT_STALL_SAMPLER: 1
+          DEBUG: pw:browser*
+        run: |
+          (while sleep 0.2; do echo "$(date +%s%3N) $(ps -eo pid=,rss=,comm= | grep -E 'WebProces' | tr -s ' ' | tr '\n' ';')"; done) > memory-suite.log &
+          poller=$!
+          vp test --project "browser (webkit)" > suite.log 2>&1 || true
+          kill "$poller"
+          grep -E 'FAIL|Tests |Duration' suite.log | head -50
+
+      - name: Upload logs
+        if: always()
+        uses: actions/upload-artifact@cf430e030ddbb5b0abf93d22962f4752f3646cd9 # v7.0.2
+        with:
+          name: webkit-stall-${{ matrix.run }}
+          path: '*.log'
diff --git a/test/browser/zzStall.experiment.tsx b/test/browser/zzStall.experiment.tsx
new file mode 100644
index 0000000000..60fac16a24
--- /dev/null
+++ b/test/browser/zzStall.experiment.tsx
@@ -0,0 +1,51 @@
+// TEMPORARY: WebKit stall experiment, does the page stall while idle after rendering a million rows?
+import { page } from 'vite-plus/test/browser';
+
+import { TreeDataGrid } from '../../src';
+
+interface Row {
+  id: number;
+  country: string;
+  year: number;
+}
+
+const columns = [
+  { key: 'country', name: 'Country' },
+  { key: 'id', name: 'Id' }
+];
+
+async function renderRows(count: number) {
+  const rows = Array.from({ length: count }, (_, id): Row => ({ id, country: 'USA', year: 2020 }));
+
+  await page.render(
+    <TreeDataGrid
+      columns={columns}
+      rows={rows}
+      groupBy={['country']}
+      expandedGroupIds={new Set(['USA'])}
+      onExpandedGroupIdsChange={() => {}}
+    />
+  );
+
+  await expect.element(page.getTreeGrid()).toHaveAttribute('aria-rowcount', String(count + 2));
+  await expect.element(page.getCell({ name: '0', exact: true })).toBeVisible();
+}
+
+// vitest-browser-react unmounts the grid before the idle test starts
+async function idle(seconds: number) {
+  for (let i = 0; i < seconds; i++) {
+    await new Promise((resolve) => {
+      setTimeout(resolve, 1000);
+    });
+  }
+}
+
+describe.each([1_000_000, 100_000])('%i rows', (count) => {
+  it('render', async () => {
+    await renderRows(count);
+  });
+
+  it('idle for 65s', { timeout: 90_000 }, async () => {
+    await idle(65);
+  });
+});
diff --git a/test/zzStallSampler.ts b/test/zzStallSampler.ts
new file mode 100644
index 0000000000..36597868fd
--- /dev/null
+++ b/test/zzStallSampler.ts
@@ -0,0 +1,42 @@
+// TEMPORARY: WebKit stall experiment, records the page's main-thread stalls and test timings
+// with wall-clock times, so stalls can be compared across pages and with memory samples
+import { server } from 'vite-plus/test/browser';
+
+const INTERVAL = 5;
+const THRESHOLD = 100;
+
+let file = '';
+let currentTest = '(before the first test)';
+let testStart = 0;
+let tests: { name: string; start: number; end: number }[] = [];
+let stalls: { start: number; end: number; test: string }[] = [];
+let last = performance.now();
+
+setInterval(() => {
+  const now = performance.now();
+  if (now - last - INTERVAL > THRESHOLD) {
+    const end = Date.now();
+    stalls.push({ start: Math.round(end - (now - last)), end, test: currentTest });
+  }
+  last = now;
+}, INTERVAL);
+
+// registered before the other setup files' hooks, so it runs first
+beforeEach(({ task }) => {
+  file = task.file.name;
+  currentTest = task.name;
+  testStart = Date.now();
+});
+
+afterEach(({ task }) => {
+  tests.push({ name: task.name, start: testStart, end: Date.now() });
+  currentTest = `(after ${task.name})`;
+});
+
+// one log per file, so the experiment doesn't flood the console
+afterAll(() => {
+  // oxlint-disable-next-line eslint/no-console
+  console.info(`STALLS ${JSON.stringify({ browser: server.browser, file, tests, stalls })}`);
+  tests = [];
+  stalls = [];
+});
diff --git a/vite.config.ts b/vite.config.ts
index 4fe6cec15b..c95dd0a3b7 100644
--- a/vite.config.ts
+++ b/vite.config.ts
@@ -311,9 +311,20 @@ export default defineConfig({
       {
         test: {
           name: 'browser',
-          include: ['browser/**/*.test.*', ...(isCI ? ['visual/*.test.*'] : [])],
+          include: [
+            'browser/**/*.test.*',
+            ...(isCI ? ['visual/*.test.*'] : []),
+            // TEMPORARY: WebKit stall experiment
+            ...(process.env.WEBKIT_STALL_IDLE === '1' ? ['browser/zzStall.experiment.tsx'] : [])
+          ],
           browser: { enabled: true },
-          setupFiles: ['test/browser/styles.css', 'test/setupBrowser.ts', 'test/failOnConsole.ts']
+          setupFiles: [
+            // TEMPORARY: WebKit stall experiment
+            ...(process.env.WEBKIT_STALL_SAMPLER === '1' ? ['test/zzStallSampler.ts'] : []),
+            'test/browser/styles.css',
+            'test/setupBrowser.ts',
+            'test/failOnConsole.ts'
+          ]
         }
       },
       {