In addition, I found the distinctions between what is confined to a single machine and what is properly clusterable (without external applications) to be lacking. For example, knowing that conferences don't replicate to external databases, nor do registrations when using mod_verto, etc. It can be easy to head down a path of getting everything working on a single machine and then realizing you have to change a lot to cluster it. In fact, generally speaking HA support seems lacking. Would be great to see more official FS documentation on example cluster setups with Kamailio/others.

A conference is very ephemeral in the concept of clustering an failover, We are working on the HA documentation, But those are highly commercialized use cases, The companies that tend to figure it out never share it back to the community, I've been working with Giovanni to correct this.  :)  In fact we do have a solution thats basically copy and paste.