a good usage scenario is that the user might want to trigger the sampling on a particular stage of a Spark job