Yes, about all you can do is check if a line is short some fields, and if it is and the next line is also short, put them together. But this is further complicated by the possibility that the newlines could be inside a quoted field, in which case the CSV parsing will choke and you have to figure out a much more complicated way of counting fields. And the file could theoretically not have a header, or the header might not have names for all the fields, in which case figuring out how many fields the file -should- have is difficult as well. And what if the field with the newline is at the very end of the record and unquoted? It'll look like that record is fine and the problem starts on the next record.
This is one of those problems I wish I had an easy answer for as well. I've run into it a number of times while doing data imports for one of my clients.