When reading a charachter type BLOB as LONG(W)VARCHAR, no/wrong charset conversion is performed.
When char BLOB is bound to SQL_C_WCHAR, mbstowcs conversion cannot be applied to each chuck returned from the DB seperately, but must be done in advance for the complete BLOB. Imagine i.e. a 3-byte UTF8 sequence cut split over 2 chunks. Impossible to convert correctly, therefore need to convert all at once in advance.