For example, the 3-Node Replset (all feature flags) variant should be running the same tests as the non-feature-flag version of that variant. 

This is important so that we can track that features being built under a feature flag are not already regressing performance compared to master.

This will speed up current-day development as well. For example, I am tracking a regression in load_test_high_value that is currently not run in feature flag variants, so I'm unable to quickly verify if a project being built under feature flag will improve this regression.