Is prr constant throughout different samples, or do we scale it with the value of the tentative contribution?

@ndiamant Typically it's just a constant parameter (though choosing this parameter is an interesting question...)