Won’t you need to retry more than 300,000 times for 100,000 original processes? Unless of course the retry somehow has a 100% success rate.

The article makes a good point though.