The goal is to fully automate the test run, in order to avoid having to hire a human to babysit the test execution and spend time rerunning tests because of spurious errors. To do so, the tests need to be repeatable without any spurious errors. Puppeteer is a complex system – there is a Node.js module communicating with a Chromium instance running Headless in the background – and it seems easy for timing issues to cause a spurious error.