Fewest trials per arm at which all-fail against all-pass reaches alpha.
The smallest trials per arm, from the observed failure rate, whose chance of reaching alpha against a control that never fails is at least targetPower.
P(Bin(n, rate) >= k) in micro-units, by a dynamic program in picos with each product floored, so the value is at most 1 micro-unit low.
P(Bin(n, rate) >= k)
Laplace's rule, (failures + 1) / (runs + 2) in micro-units: never exactly 0 or 1.
(failures + 1) / (runs + 2)
Fewest treatment failures out of n, with a clean control of n, that reach alpha.