It's a hard problem, and the Postgres team's careful approach to engineering means that they are solving it slowly, in comparison to the current fashion for shi

It's a hard problem, and the Postgres team's careful approach to engineering means that they are solving it slowly, in comparison to the current fashion for ship first, patch later. They bring stuff in piece by piece, eg. in 9.4 they added logical changeset extraction/logical log streaming replication, which is another step down the road to clustering. On top of this they are building bi-directional replication. And so it continues. This is the same reason they haven't added upsert or merge yet.