Limitations

Values are strings. There is no mapping between rows and objects and no type conversion: parse the values yourself, with the invariant culture.
Asynchronous only. Reading and writing are async methods. There is no synchronous API and no CancellationToken.
No trimming and no comments. Spaces around a value are part of the value, and a line that starts with # is a row like any other.
No detection. The reader does not guess the separator, the quote character, whether the first row is a header or the encoding. Set Separator, Quote and HasHeaderRow, and pass a TextReader that decodes the right encoding.
Case-sensitive column names. row["name"] does not find the column Name; see Rows with a header.
No validation. Rows can have any number of values. Check row.Values.Count when the data must have a fixed number of columns.