Why? Very few products need sharding, let alone multi-master. Sure a popular social media platform would, but most development in the world is for small to medium scale line-of-business apps. Postgres is _fantastic_ for these.
This is the key point we see - most applications just do not generate write traffic that is beyond what a recent release of Postgres can handle.
I manage a large sized Postgres farm with 100s of instances, and there has been one case where we need multi-master, and I went with Galera cluster for MariaDB. You can shard using the citus extension for Postgres.
Depending on requirements, there are an increasing number of options for "active-active" Postgres deployments. A colleague wrote this on a federated active-active configuration on Kubernetes: https://info.crunchydata.com/blog/active-active-postgres-fed...
Postgres supports table partitioning and foreign data wrappers (used for accessing remote SQL databases) which can be used to set up sharding as described in the postgres docs.
It's because most developers don't need multi-primary databases, which have downsides too.
Was using aws pg service and wanted our prod@aws synced to dev/backup@localhost and found Bucardo[0] to do a great job.
"Bucardo is an asynchronous PostgreSQL replication system, allowing for both multi-master and multi-slave operations"