I download data incrementally every hour from some sources by extracting all records which have been updated/created in the last hour. This means that I end up with very small p
