Roughly 50% chance of messing up with data in #3. There is the possibility of reducing a bit this chance at the cost of wasting more space by using error-corre

Roughly 50% chance of messing up with data in #3. There is the possibility of reducing a bit this chance at the cost of wasting more space by using error-correction, e.g. with RAID, if so desired. This is explained in the paper and in the FAQ and the README.