It seems in the conversion from phantomjs to headless Chromium, the fail.timeout event is no longer emitted, and there was no replacement for ensuring that page that is stuck with errors eventually leads to a failure.

Instead, the task just hangs forever. This is further made worse now that projects on GitHub have largely switched from Travis to GitHub Actions, which usually does not show build output while a build is in progress. As such, one has to wait for 10 or 30 minutes for the overall build to timeout (or cancel the parent workflow), in order to find out which test was last reported as running.