The distinction between query.run(conn) and conn.run(query) is very fine indeed. There has been much discussion about this and there are both solutions have their merits. At one point we even supported both but decided the simplicity of having only one solution merited dropping the other.
Creating a new connection involves opening a TCP connection and sending a message to validate the driver to the server, altogether a few round trips. There is a small bit of per connection state stored on the server but since RethinkDB uses a custom coroutine implementation for concurrency support this does not amount to the overhead of an independent OS thread for each connection.
Invoking the query with a specific connection object is thread safe. There is a feature designed to help REPL users that stores the last connection in global state that is slated for removal in the upcoming release (1.4) that is obviously not re-entrant.
It would make sense to have connection pooling, especially in the python driver where connections block on requests. This is an idea we're exploring but is lower down on the priority list. As it is a fully client side feature there is nothing stopping 3rd party driver developers from implementing a solution though the official drivers will have to wait for other priorities.
I don't know if I'd call it an issue with the API -- there is a pretty active debate going on about it: https://github.com/rethinkdb/rethinkdb/issues/256 Please chime in there with your POV, it would really help!
how expensive is the creation of a new connection?
Connections aren't thread safe on the clients. It's very efficient to open a connection to the server, but it still requires a TCP handshake. This is a good low-level API -- we'll build a connection pool on top of that soon (I just opened an issue for it here -- https://github.com/rethinkdb/rethinkdb/issues/281). People seem to want one because they're used to one, but it hasn't really been necessary in our testing.
g.rdb_conn.table('todos').get(todo_id).run()
On the other hand, this form ties the request to the connection, which neither the current API or my original suggestion do.
Also, you are opening and closing DB connections with each HTTP request. I imagine this is for the purpose of simplifying the example. However, it does make me curious: how expensive is the creation of a new connection? Are connections thread-safe? Would it make sense to have connection pooling?