I have used the odbc drivers for csv previously. I agree with your assessment, never again. They are unforgiving in what they except and not a robust solution at all. I spent months cleaning up edge cases the driver should have been able to handle.
The built-in tooling for CSV in PowerShell is excellent (provided you don't care about details like if numbers are quoted or not) - I've been using that for csv-related ETL tasks.
