Raised up in #25059 something is not right with the baseline measurements given the #s are the same or smaller for these large cases than the small cases. Likely that is more of a harness issue. Could because it is measuring the test runner process instead of the CLI process. Need investigation.