I have a client with about 5 GIS authors. Their primary job is to create basic web maps and story maps for their clients. At any given moment, there might be 25-50 people viewing the end product. All in all, I don't view this as a highly stressed system (no geoprocessing, analytics, etc).
They also want the system to be high availability with disaster recovery using a series of virtual machines with flexible resources.
Is this a good use-case for just doing a single-machine deployment in production? I feel that this would limit bandwidth usage between HA machines and improve performance. The only "downside" is that you can't scale out--but I can always add storage, and I can't imagine them burning through all available cores with a handful of active story maps.
From Esri's documentation:
The [single-machine] deployment scenario described below is straightforward to set up, maintain, and upgrade. It can support a sandbox environment for development and testing, but it is also a valid (and in some cases ideal) configuration for some production environments.