Should you retry failed Selenium tests automatically?

A retry hides a failure on the first attempt. That is acceptable when the cause is a known transient such as a Grid node restart, but dangerous when it masks a real race in the product or the test. The default should be to fix the cause.

public class RetryAnalyzer implements IRetryAnalyzer {
    private int attempts = 0;
    private static final int MAX_ATTEMPTS = 1;

    @Override
    public boolean retry(ITestResult result) {
        if (attempts < MAX_ATTEMPTS) {
            attempts++;
            return true;
        }
        return false;
    }
}

@Test(retryAnalyzer = RetryAnalyzer.class)
public void checkoutWorks() {
    new CheckoutPage(driver()).completeOrder();
}

• Retry the smallest scope possible, ideally a known flaky test rather than every test
• Keep the retry count low; repeated retries multiply runtime and hide systemic issues
• Record that a test passed on retry and report it separately from a clean pass
• Quarantine tests that need constant retries and assign an owner
• Do not retry assertion failures that indicate deterministic product bugs
• Ensure retries do not leak data or duplicate side effects

A retry analyzer may be instantiated or reused in ways that make a simple counter unsafe under parallelism. Verify the runner's behaviour or use a thread-safe implementation. The retry decision should be visible in the report so nobody mistakes a retried pass for a healthy test.