It would be great if instead of using a locally installed instance of Postgres for the relational datastore on a server, one could opt to use a serverless offering like Azure Database for PostgreSQL | Microsoft Azure or Amazon RDS for PostgreSQL – Amazon Web Services (AWS) , Particularly (probably only) while deploying ArcGIS Enterprise in the respective clouds.
As it stands, the HA capabilities of ArcGIS Relational Datastore and the Portal database are rather weak, supporting only Active/Passive failover and the failover process is stateful and prone to failure (there is no witness!).
I would much rather use a serverless solution in the cloud. This could be done by installing the relational datastore component somewhere (say on the hosting site machine, but deploying the database managed by the data store service elsewhere), say Azure Database for PostgreSQL. I feel like it would work the same, but jsut offload the actual RDBMS to the cloud platform instead of a local install.
Not only would this save money for customers (less hardware costs.. serverless DBs are always cheaper) but it would be far more reliable and scalable.
With the move to containerization in the Kubernetes side of things, this should be easy enough to do and would be the ideal solution for running that component even in a Kubernetes world.