diff --git a/.github/workflows/e2e.yml b/.github/workflows/e2e.yml
index 2f05a8f..df9440f 100644
--- a/.github/workflows/e2e.yml
+++ b/.github/workflows/e2e.yml
@@ -29,10 +29,34 @@ on:
 
 jobs:
   e2e:
-    name: E2E integration tests (MSW mocked)
+    name: E2E (${{ matrix.project }})
     runs-on: ubuntu-latest
     timeout-minutes: 30
 
+    # fail-fast: false is the point of this matrix — if WebKit breaks we
+    # still want Chromium/Firefox/mobile results in the same run, each
+    # clearly attributed to its own job rather than one aggregate pass/fail.
+    strategy:
+      fail-fast: false
+      matrix:
+        include:
+          # `project` maps 1:1 to a project name in e2e/playwright.config.ts
+          # and is passed to `--project=<name>`.
+          # `browser` is the underlying engine Playwright needs installed —
+          # the two mobile projects emulate a device on top of an existing
+          # desktop engine, so they reuse that engine's binary rather than
+          # needing a separate download.
+          - project: chromium
+            browser: chromium
+          - project: firefox
+            browser: firefox
+          - project: webkit
+            browser: webkit
+          - project: mobile-chrome
+            browser: chromium
+          - project: mobile-safari
+            browser: webkit
+
     defaults:
       run:
         working-directory: e2e
@@ -66,14 +90,14 @@ jobs:
         run: npm ci
         working-directory: e2e
 
-      - name: Install Playwright browsers
-        run: npx playwright install --with-deps chromium
+      - name: Install Playwright browser (${{ matrix.browser }})
+        run: npx playwright install --with-deps ${{ matrix.browser }}
         working-directory: e2e
 
       # ── Run tests ────────────────────────────────────────────────────────────
 
-      - name: Run E2E integration tests
-        run: npx playwright test --reporter=github
+      - name: Run E2E tests — ${{ matrix.project }}
+        run: npx playwright test --project=${{ matrix.project }} --reporter=github
         working-directory: e2e
         env:
           CI: true
@@ -86,12 +110,14 @@ jobs:
           E2E_API_BASE_URL: ${{ vars.E2E_API_BASE_URL || '' }}
 
       # ── Artifacts ────────────────────────────────────────────────────────────
+      # Artifact names include the project so a reviewer can jump straight
+      # to (e.g.) the mobile-safari report without guessing which run it was.
 
       - name: Upload Playwright report
         if: always()
         uses: actions/upload-artifact@v4
         with:
-          name: playwright-report-${{ github.run_id }}
+          name: playwright-report-${{ matrix.project }}-${{ github.run_id }}
           path: |
             frontend/test-results/
             e2e/playwright-report/
@@ -101,6 +127,6 @@ jobs:
         if: failure()
         uses: actions/upload-artifact@v4
         with:
-          name: playwright-traces-${{ github.run_id }}
+          name: playwright-traces-${{ matrix.project }}-${{ github.run_id }}
           path: frontend/test-results/**/*.zip
-          retention-days: 7
+          retention-days: 7
\ No newline at end of file
diff --git a/e2e/playwright.config.ts b/e2e/playwright.config.ts
index 7b9cb2b..3458bf9 100644
--- a/e2e/playwright.config.ts
+++ b/e2e/playwright.config.ts
@@ -11,6 +11,11 @@
  *   cd e2e
  *   npx playwright test
  *
+ * ## Running a single browser engine locally
+ *
+ *   npx playwright test --project=firefox
+ *   npx playwright test --project=mobile-safari
+ *
  * ## Running against a real bridgelet-sdk instance
  *
  *   E2E_API_BASE_URL=http://localhost:3001 \
@@ -47,8 +52,10 @@ export default defineConfig({
   retries: process.env['CI'] ? 2 : 0,
 
   /**
-   * GitHub Actions reporter annotates the PR check directly;
-   * the list reporter is easier to read in a local terminal.
+   * GitHub Actions reporter annotates the PR check directly, and includes
+   * the project (browser) name in each annotation — this is what gives us
+   * per-engine failure attribution in the CI output. The list reporter is
+   * easier to read in a local terminal.
    */
   reporter: process.env['CI'] ? 'github' : 'list',
 
@@ -91,14 +98,40 @@ export default defineConfig({
   },
 
   /**
-   * Only run Chromium in CI to keep the scheduled job fast.
-   * Add more projects locally if cross-browser coverage is needed.
+   * Cross-browser + mobile-viewport matrix.
+   *
+   * Desktop engines cover the three major rendering engines (Chromium,
+   * Gecko, WebKit). The mobile projects emulate real device viewports,
+   * touch input, and UA strings — claim links are frequently opened from
+   * SMS/mail apps on a phone, so a desktop-viewport-only pass would miss
+   * layout and touch-target regressions on that path.
+   *
+   * Each project name is what CI's matrix strategy passes to
+   * `--project=<name>`, and what shows up in the GitHub Actions job name
+   * and Playwright's `--reporter=github` annotations — so a failure is
+   * always attributed to a specific engine/device, not just "e2e failed".
    */
   projects: [
     {
       name: 'chromium',
       use: { ...devices['Desktop Chrome'] },
     },
+    {
+      name: 'firefox',
+      use: { ...devices['Desktop Firefox'] },
+    },
+    {
+      name: 'webkit',
+      use: { ...devices['Desktop Safari'] },
+    },
+    {
+      name: 'mobile-chrome',
+      use: { ...devices['Pixel 7'] },
+    },
+    {
+      name: 'mobile-safari',
+      use: { ...devices['iPhone 14'] },
+    },
   ],
 
   /**
@@ -122,4 +155,4 @@ export default defineConfig({
   },
 
   outputDir: '../frontend/test-results',
-});
+});
\ No newline at end of file
diff --git a/frontend/components/dev-toolbar.tsx b/frontend/components/dev-toolbar.tsx
index a3afbb0..a92cce1 100644
--- a/frontend/components/dev-toolbar.tsx
+++ b/frontend/components/dev-toolbar.tsx
@@ -54,6 +54,18 @@ export function DevToolbar() {
   // Avoid rendering until hydrated so the radio group shows the correct value.
   if (!mounted) return null;
 
+  // Skip rendering under browser automation (Playwright, Selenium, Puppeteer,
+  // etc. all set navigator.webdriver = true; real users never do). This
+  // toolbar is dev-only tooling for manually switching MSW scenarios — the
+  // E2E suite selects its scenario via the URL token instead
+  // (e.g. /claim/expired-e2e-test-token), so the toolbar isn't needed
+  // during automated runs. Without this check, the fixed bottom-4 right-4
+  // panel renders during E2E too (next dev still has
+  // NODE_ENV === 'development'), and on the shorter mobile viewports it
+  // overlaps and intercepts clicks on real submit buttons (e.g. "Review
+  // Payment" / "Continue"), causing spurious mobile-only E2E failures.
+  if (typeof navigator !== 'undefined' && navigator.webdriver) return null;
+
   if (collapsed) {
     return (
       <button
@@ -143,4 +155,4 @@ export function DevToolbar() {
       <p className="mt-2 text-[10px] text-slate-400">Dev only · persisted to sessionStorage</p>
     </aside>
   );
-}
+}
\ No newline at end of file