How are server failures handled?
If we can’t connect to a server or there is a problem (anything other than a 200 OK or Accepted response) we take corrective action. The problematic server is initially excluded from all ping requests for 5 minutes. We then retry with delays based on a factor of 5 mins (5/10/20/35/55/80/110). When 15 failures are reached (about 48 hours) the server is automatically permanently deactivated.