Instead of creating smaller CSVs, I'd probably try using the original CSV but doing rows in chunks of 500. You could create a data store to store a value that represents which row to start on, and increment that value by 500 every time the scenario runs. First time it runs, start on row 1. Second ti...