For JSON, the format we choose is not to have one single object per file, but one JSON object per record, one record per line. One of the advantages of this format is that it is easy to split a JSON file to do parallel processing. This approach is not specific to Dremio but used by multiple data analytics frameworks.
For CSV file, I believe Dremio will put the field under double quotes (as the standard mandates) if the field contains a comma. Isn’t it not the case, or do you prefer a TSV file format as you are not using a CSV parsing library to process the content?