Particularly problematic with this is that a human cannot necessarily determine with 100% confidence where the error is. If you have a CSV where every row has 5 entries, and you come upon this row:

> 100,1,200,300,10,100

You suspect that an in-field comma is causing a split where there shouldn't be, but who's to say which number got split up? 1200 into 1,200, or was it 200300 to 200,300, or 10100 to 10,100?