Sometimes I have a test going with 4 or 5 variations and notice that 1 is dramatically underperforming the rest. Would be great if we could drop one (or more) from the test but keep it running with the remaining variations.