We have several tickets for tests that fail non-deterministically in Evergreen.

See the flaky-tests label for a list of all such tests.

To more easily diagnose these, let's consider:

• running these tests "n" times on a dedicated variant.
• enabling trace logging for that variant to get verbose log output
  • take into consideration that trace logging could be verbose. E.g. if we run AWS tests or Atlas auth tests with tracing, those should not place trace logging output anywhere public.
• scheduling this variant at a lower cadence, perhaps once a week, and have it not be a part of patch builds