Re: How QlikSense balance the load internally in multi node environment.
Regarding the scheduling algorithm being round robin, there is a pretty strong caveat on that.
It seems that the node allocated to a user when they log on becomes the node used for any subsequent continuous sessions. That means that if you logon to Qlik Sense in the morning - when it is all quiet and you go to the central node, if you keep Qlik Sense open all day, and open a bunch of other apps etc. during the day without being timed out, all of that activity happens on the node you originally logged on to - regardless of the fact that it might be under serious load from other logons.
So it only round robins new connections. Which means that you probably want to think seriously about separating out the different types of behaviour for load balancing.