Does optimising the tests towards one architecture seem like a fair way of testing?