Right now, if any column name/field in CSV records to combine via Semigroup instance is shared between the records, the Semigroup[NamedRow] Instance will complain. But this could be relaxed, where tolerance of overlapping fields is given if the values are identical.