Selenium

We use Selenium where a suite has to outlive the team that wrote it. Tests are structured as page objects so locators live in one place and a UI change is a single edit rather than a sweep through fifty files. We standardise on explicit waits and drop implicit ones entirely, because mixing the two is the most common source of flake we inherit. Runs are parallelised through TestNG and distributed across Grid or BrowserStack depending on whether the client wants to own infrastructure. Everything is wired into CI so the suite runs on every merge, not on request. When we take over an existing Selenium estate — which is most of the time — the first pass is usually deleting: quarantining tests that fail intermittently, removing duplicated coverage, and getting the run time low enough that developers stop ignoring the results.