Quite the opposite. The running time of sklearn's or xgb's training procedure is not the main obstacle to retraining daily (not even close). In some of industries retraining has to go through rigorous manual QA. And you need a solid ML pipeline to begin with.