There's logic behind the complicated rules. The idea is that you can't cause the calculated availability to go down by generating lots of requests when the service is down.