Came here to post the same. This can be used in very interesting ways, either for quick ad-hoc queries or for import/export functionality (ETL), where the core of your data is in InnoDB but you keep a CSV version for one of the phases of your processing, to interface with other tools.

This may looks like a dirty hack compared to a clean API, but if you have performance considerations or need to interface with some legacy thing that happens to understand CSV, this can be a good choice.