I'm sorry you've needed failover. Since we've opened the pool, we've kept 99.84% uptime. That's may strike you as good reliability, but I know we can do better. Over half of that time was spent in our last major deploy on June 12, when we switched to a system that lets us hot deploy with no downtime. Since then we've kept 99.93% uptime, with only one bad deploy taking us down for 12m on June 18.
We've just started tracking response time as well, so I'm looking forward to telling you about that in a month or so.