* when management server dies and notifies other management servers about this, the running management server has to cleanup host_transfer records belonging to the died management server
* issue agent load balancing task only when agent load (number of connected agents in the system) exceeds "agent.load.threshold" - 70% by default