Functionality for a failover does not exist as part of the SMP, but you can plan for this sort of functionality as part of your disaster planning. If computers are reporting to NS-A, and NS-A fails, you can create a CNAME DNS entry for NS-A so that it resolves to the IP address of NS-B. As a result, clients will talk to the correct NS. If you're replicating resources, they will already be known to NS-B.
You are correct that hierarchy and replication can be used to plan for recovery, but are not backup plans and are not hot failovers. That's because the agents are still reporting to the dead NS and something needs to tell them to check in with the correct NS. The DNS CNAME entry can do this.
The best plan is to properly back up your database and your NS, and to store the MSI's from your NS in a safe location. With a server ready to be built--or even better, a VM template ready--provision the server, execute all the stored .msi's, you're now running the exact same versions as your old NS and you just need to connect to the SQL database.