The problem I've noticed is that if I append the file with each of the 100k runs (one at a time), it can happen that two threads try to save to the file at the same time and some row (s) end up empty.