I should have included the context that my notes pertain somewhat more to Postgres than other database systems; "USE" may work for MSSQL (and others) where the

I should have included the context that my notes pertain somewhat more to Postgres than other database systems; "USE" may work for MSSQL (and others) where the database is treated as more of a namespace concept, whereas Postgres requires a new connection per database. If you've got 1,000 databases on a Postgres cluster and are limited to 1,000 connections, there's going to be a tradeoff in connection setup latency vs. connection concurrency vs. efficient resource use. Or, spend time writing a dynamically weighted set of pools? Oof. A parameterized query like "SELECT * FROM post_categories WHERE tenant_id = ?" would very much be cacheable by the query planner. Stuff like "does tenant_id have an index? is it worth using the index based on its histogram? are those index pages already in memory?" etc.