Over the last few months, I’ve seen WPT tests fail when I enable multiple runs. The typical issue is that websites have implemented some kind of rate control to prevent attacks or bad bots. WPT tests are being categorized as something malicious and the requests are blocked.
My suggestion is that WPT should provide an option to introduce delay between multiple test runs. This way, the tests will generally escape the rate limits and continue with the tests.
Please let me know if this sounds feasible.