Thanks for the suggestion. It can indeed make sense to start a WFO test several times with different starting points that are shifted over a WFO test period. This multiple test would catch such problems as you've described. I'll do some experiments with this method.