Distribute health checks over interval

[jbrendel]

For every host in the cluster a thread is started for the health check. Many of those threads finish quickly. However, all threads are spawned at the same time. They should instead be started in a staggered fashion.