Because Enterprise Server “stays in the middle” of NX session in the current realisation and is a point of failure. Maybe clusterisation helps with it, but, as I understand, anyway all nx sessions will be disconnected when the cluster switch (or not?).
The high-availability feature of NoMachine monitors the servers in an active-passive set up. Check this document for more details:
Sessions on the NoMachine nodes are preserved in case of failover and can be reconnected.