There was three minutes of downtime for one server which was distributed in three instances when one of our switches switched to a different subnet.... and that's thanks to the ARP tables.
Just so you know we have multiple servers and this update had no effect on anyone as we use HA clustering.
