It sounds like a good approach and it should be setup so only one server is writing to the files and all other ask that server for the result (yes, single point of failure).
I wrote a blog post about this since that is the only way to make Episerver Search work in a Azure enviornment, you can read it here:
I would expect this is a fairly easy thing to configure but I've not been able to find out how to do it so hopefully someone can offer some pointers.
We have a load balanced environment with a number of presentation servers and a number of edit servers. Even though we are using Find the client still uses the search in the CMS on the edit servers in order to locate pages within the CMS. As this is not necessary on the presentation servers we would like to turn off the indexing so that only the CMS servers are performing indexing and calling the indexing service.
The reason behind this is to try and reduce the load on the presentation servers.