One way of doing this would be to issue a few requests on some host that is known to have a near-100% uptime and only start the actual probing if the requests succeed.
One way of doing this would be to issue a few requests on some host that is known to have a near-100% uptime and only start the actual probing if the requests succeed.