our internal approach has been to have timeouts in CI, since they address multiple failure cases, including tests/ivy/network flakiness