The CSV parsing library used is papaparse (https://www.papaparse.com/). Maybe something in there?
Try smaller chunks of the file to see if it makes it easier to spot, see what happened to the docs immediately before or after the gaps.