If the source data's schema has changed, or changes have been made to the source configuration since the last load, then you should choose "Restage and reload".
This option is a "completely fresh" initial load. Precog ignores any prior information about the dataset, and pulls everything from the API again from scratch. The original schema is not maintained. Precog inferes the schema entirely from the newly fetched data. Precog writes the data to the database as if this were an initial load.
Why - there is an issue in the data or the schema and the customer is not concerned about the schema changing.