We are all overwhelmed with email alerts on dataset failures, right?
How often do you open the dataset only to find that it was a single random failure, and has since run successfully?
I think it would help to have a setting where the dataset/dataflow failure alerts only trigger when there have been multiple consecutive failures.