These are good points, but from what I'm seeing it's now more in vogue to use sharding (or other favorite term for same) to split your data by user across different DBs. That way you can still scale horizontally.
Is there an open source sharding solution for Postgres?
Sharding is where you split the data across multiple servers, based on some identifying category like users, functional area, etc.
An Unorthodox Approach to Database Design : The Coming of the Shard [http://news.ycombinator.com/item?id=38407]
Partitioning vs. Federation vs. Sharding [http://www.lethargy.org/~jesus/archives/95-Partitioning-vs.-...]