Once #7419 has been implemented we'll need to have the ability to decide whether to read from a replicated cluster or a single node.
We'll use DNS for swapping reads. We can have a higher level SRV record that points to another SRV record and swap that over via OMDB. Then we can use a policy to make the change in reconfigurator.