diff --git a/docker-compose.e2e.yml b/docker-compose.e2e.yml
index 681136e..cd1ab66 100644
--- a/docker-compose.e2e.yml
+++ b/docker-compose.e2e.yml
@@ -90,7 +90,7 @@ services:
         done;
         echo 'Waiting for backend...';
         for i in \$(seq 1 120); do
-          curl -sf http://backend:8080/api/vehicles/ZZZ999 > /dev/null && break;
+          curl -s -o /dev/null http://backend:8080/api/vehicles/ZZZ999 && break;
           sleep 1;
         done;
         echo 'Waiting for frontend...';
diff --git a/frontend/playwright.config.ts b/frontend/playwright.config.ts
index 8613604..7daf7a5 100644
--- a/frontend/playwright.config.ts
+++ b/frontend/playwright.config.ts
@@ -5,7 +5,14 @@ const isCI = !!process.env.PLAYWRIGHT_BASE_URL
 export default defineConfig({
   testDir: './e2e',
   timeout: 30_000,
-  retries: 0,
+  // CI flakes: the E2E stack runs 4 parallel Playwright workers against a
+  // single backend (Spring Boot, no -Xmx cap). Under load an occasional
+  // order-creation request transiently fails, which surfaces as a spurious
+  // "navigation to /betalning/ timed out" failure unrelated to the code under
+  // test (e.g. run 103, where 2 navigation tests failed while 94/94 passed
+  // locally on the same commit). Per Playwright's guidance, retry transient
+  // failures in CI; keep retries off locally for fast feedback.
+  retries: isCI ? 2 : 0,
   use: {
     baseURL: process.env.PLAYWRIGHT_BASE_URL || 'http://localhost:3000',
     headless: true,