How to configure a Publisher cluster? Are Publisher clusters highly available?
We are designing a QlikView server configuration to be "highly available", to provide continuous service with minimal human intervention when individual components of the system fail. We also want to improve performance by implementing some sort of load balancing across multiple servers. 
In addition to providing high availability and load balancing by clustering named CAL licenses for QlikView Server (using network load balancing or AccessPoint load balancing), we would like to provide high availability for QlikView Publisher.
We have read the QlikView Server Reference Manual v10 section on Publisher Load Sharing (Clustering), http://tinyurl.com/3p3ofpv. This document (page 197) implies that multiple QlikView Distribution Services can run on several servers, sharing load according to a defined formula.
(1) The documentation on page 197 says:
In order to cluster QlikView Distribution Service, the services will need a common disk area on a NAS to save the configuration file. Add the same value to the setting <ApplicationDataFolder> in the QlikViewDistributionService.exe.config for all Distribution Services that should be clustered.
I cannot find this file anywhere. How do I configure the QlikView Distribution Service(s) to look to this common disk area? Does common disk area simply mean a directory?
(2) If multiple Publishers are clustered, how do they become aware of each other? Thru this common disk area, or the config file mentioned?
(3) If we can cluster several Publishers, the documentation implies they will share load. Will the Publisher cluster also provide high availability? For example, what happens if we have three Publishers in the cluster and one Publisher goes down or otherwise becomes unavailable? Will the remaining two Publishers continue to function and re-distribute the load between them? (This would provide high availability, which is our goal.)
 We have read the QlikView white paper, "Cluster QlikView Servers for Resilience and Horizontal Scalability", http://tinyurl.com/3jtg6ps. This paper (page 3, et al) implies that load balancing can be implemented with a single QlikView AccessPoint routing requests to multiple QlikView Servers, or with network load balancing (HW or SW based) distributing requests to multiple QlikView AccessPoints.
By the way, Figure 2 on page 3 implies that the multiple AccessPoints should be aware of all nodes in the cluster; whenever we do that, the browser client crashes when you try to open the same document twice in the SAME IE browser process. We did set the Root Folders on all QlikView Servers to the same network folder. The diagram appears to be wrong; i.e. I think AccessPoint 1 should be serving up QVS1, only.