Comment 2 for bug 1339659

Fix in 105719 works to eventually discard messages sent via TCP to a nonresponsive rsyslog server. The best option is to use UDP to avoid this scenario. After a couple of MB of logs get generated, all old messages get collected and sent to Fuel Master. This will repeat indefinitely, though, distributing logs on and off in spurts. It is still a better option than completely broken logging when 2nd server is broken.