Why would the S3 bucket be so much more massive than the data currently being stored in it? We are loading data into the tables using kafka routine loads.