To improve developer productivity, this is a proposal to move browser-based tests (Selenium and QUnit) to a non-voting pipeline, so that flaky browser tests cannot block merge or verification +2 of patches.
If this step is taken, then it is up to the reviewing developer or the developer who gave a +2 to inspect the console output and judge for themselves whether a possible regression is involved with the patch. While not ideal, this seems much more preferable to the current status where patches are regularly blocked on merge (after 30-60 minutes of waiting) only for a "recheck" comment to allow them to succeed (most of the time).
related tasks: