While you expanded the amount of rows that can run for a preview, what would be even better is being able to select data that is previewed, even if it's something like most recent 400K rows. Changes that we generally make to an ETL is due to updates in the data flowing through but it's difficult to test that logic when the ETL only tests the first set of rows or a random set of rows. I then have to create copies or Views of datasets, filtered to the rows I need to test, so I can preview that my change in logic worked before reverting back to the original datasets and rerunning the full dataflow.
