Original JSON is required if you want to do partial document update (merge) or reindex. Since logs are not updated I have used index without source for that use case.
Also, if you are always fetching all fields, getting _source instead of 100 fields using doc values or stored_fields may be faster. I haven't benchmarked it.