@nodejs/collaborators
https://ci.nodejs.org/job/node-stress-single-test/ is now available and ready to use. It can be used to help determining if a test is flaky or not.
To test a pull request, use refs/pull/99999/head in the GIT_REMOTE_REF field.
The default of 100 runs can be useful to get the error output of a test that fails frequently, or to quickly test a fix attempt before a longer test. To know with a high probability that a test is not flaky I suggest 10000 runs, because I've seen a flaky test failing about 3 out of 10000 runs. For now these are just suggestions, we can change the default and create best practices when we have more experience with this.