A data flow depends on a refreshed source dataset. How should an analyst arrange repeatable processing?

1. Give the output dataset the source connection password as its name
2. Sequence dependent refresh and preparation steps in their required order
3. Remove all saved outputs after each successful run
4. Run every step at random times to avoid contention