Open axch opened 8 years ago
Project management decision: Is this a priority?
It makes sense that we'd want to run backend-agnostic tests on multiple backends, but running slow ones on only one backend most of the time also makes sense. I don't think it's a high priority since if I understand correctly it's not leading to major resource crunch, test failures, or missing information; we're just using more compute than we need to?
Any of those that have a non-negligible false fail rate impose an additional cost, namely increasing the false fail rate of the test suite as a whole, thereby reducing our trust in it. In the glorious future the proposing of which is #549, this will manifest as needing somewhat more computation for every probabilistic test in the whole test suite in order to retain the same level of trust.
to prevent them from being redundantly rerun by e.g. the all-crashes script (that Jenkins uses).
Issue: Not sure how to mechanize maintenance of the desired state (which, of course, is why this weeding seems necessary).