The full data set is also available as a postgres dump (basically, tab-separated values with \N as NULL). This dump is sorted by source_ids, so you can bisect in it.
To obtain the dump (which we don't link here to keep crawlers from needlessly pulling 48 GB of data), run your preferred variant of:
curl -O http://vo.ari.uni-heidelberg.de/gedr3spur/dump.txt.gz
For the metadata, inspect the column metadata below in database order.