The cheapest option costs more once you price in the failure rate.

Loose evidence is not nothing, but a washout either side is the only reason to trust your own take.

Half of what people describe as tolerance is a change in the underlying condition, and the two need different answers.

You have two changes at once here. You cannot attribute anything until one of them moves.

Where does the human data stop and the extrapolation begin? That is the whole thread.