When the UMS or ICG is down for an extended period of time (as short as 8 hours or less in our experience), clients stop trying to reconnect until they are rebooted, which can only be done by the end user at this point. In a largely remote deployment, this is completely unmanageable.
We have thin clients deployed all over the nation. This makes it very difficult to manage them at scale, as we have to reach out to onsite staff to reboot them before we can reestablish the link.