I realize this would be taking things to the extreme for the sake of the one case in which the test failed, and furthermore is not exactly the kind of "feature" you would want to be promoting all that much. -- That said, if the overall performance loss is signicant enough thats its noticeable in the prefevious 4898 test cases, could it be justified to implement a commandline switch, or an internal API-based trigger to turn on the slower solution in cases when the reason for the failed test is because of this bug? Would such a solution even be possible, or would the slow down take place regardless of a commandline switch/API-based trigger?