I think you are missing some important facts, you don't really "move" an instance, when you add an instance of SQL it installs ALL of the required binaries to run SQL, these are NOT shared, a copy exists on all nodes where that instance has been added. I generally suggest you add ALL SQL instance to all nodes. You can then change the Resource Group to not use nodes that you don't want but in a pinch it allows you to put them there if needed. A cluster resource group can have node preferences which helps decide where a failover goes to. For SQL there are also available nodes, when you add a node using the SQL install when it is done the node appears in that list, if you don't want this instance to use that node uncheck it. This means that without intervention that instance will NEVER run on that node.

I am concerned by the statement "re-install", you don't re-install, you add a node to an existing sql cluster, this is absolutely important, DO NOT install a stand-alone instance. There is an option in the SQL install that handles this setup you MUST use that. If you try it any other way you are pretty much screwed.

When you failover a clustered SQL server what happens is SQL is shutdown and the resources are taken offline on the node that has them, "moved" to the node that is getting them and then brought online.

I have build a couple of three node clusters and a bunch of two node but I am getting the feeling that you have never even built a two node cluster, I would try this in a dev environment before I tried it on a production server. Once you can install a two node cluster adding a third node is basically the same as the second node. I did it with VMware and VirtualBox on a laptop recently so I know it can be done in a small environment if needed.

CEWII