Open
Description
The current methodology for saving internal dataframes from the task tracker is a bit flawed - it's mostly just writing parquet to a local path. We'd prefer the ability to save dataframes to something else - like S3 - so let's make that happen.