Usually people make sure tests run long enough... but then again we all know classical statistical tests become significant eventually...