As part of the cost-based ranker initiative, we have introduced several passthroughs which are forced to disable tests which either directly (assertions on explain) or indirectly (assert on number of keys examined, memory limit hit, etc.) that a particular plan was chosen. To not lose coverage of such test cases, we should come up with a way to detect a test is running with CBR enabled and in these cases make alternative assertions about the chosen plan - after verifying that CBR doesn't choose a worse plan.