The parquet files include metadata that is used to delete existing data. The existing data refers to the same day that the current upload is refering to. the current upload is newer and therefore existing data should be removed. This functionality is handled by prepare_parquet_s3

There is another set of tables that are created via SQL files and used by report_db_accessor_base.py and its successors. For these tables we will not use partitioning at the moment.