I just ran 111 doublecheck tests at around n=1.7M in order to ge the current error rates, but none out of 111 tests was faulty; so this is only to inform everybody that doublecheck is not an issue yet.
Banana?:banana: .
Yours H.

I seriously wouldn't worry about error rates until first pass reaches about 4M.

At that point I'd re-run about 1000 test@4M, and then 3M.