So, in general, you don't "cluster" the WSAs
You use the M to push one config to multiple boxes, and to centralize reporting so you don't have to to multiple boxes to track down how much surfing in the wrong places someone has been doing.
In transparent mode, you use WCCP to handle load balance/failover.
In explicit mode (eg you tell every machine where the web proxy is), you'd use the PAC file config, or an external load balancer to get each browser/app to know where to go if the proxy fell over...
In 8.5, they've implemented a failover mechanism for explicit proxy configurations using CARP... (under Network/High Availability in the GUI)