In my application we can't afford to have the agents down for long time (without being noticed). I am looking for an option where we can set up the monitoring on agents so that when an agent fails due to an exception we get an immediate notification.
We also have something like master-for-agents to monitor the agents defined through agents rules and agent schedule data instances. By default, the Master-for-Agents master agent awakes every ten minutes. Your site can establish another value in the prconfig.xml file setting:
<env name="agents/newqueuewakeup" value="NNN" /> where NNN is in seconds.