Interesting this is a different solution than if you can't go back after interviewing someone (which is interview 1/e people then take the best after that, which gives you the best 1/e times)