Interesting. I've been monitoring my clusters with a custom Python script that grabs a lot of the data from the pg_stat_ tables, along with the current xlog location and table / database sizes, at regular intervals and saves it in a Graphite database, which I then examine with Graph Explorer [1] and the normal Graphite frontend.

This looks like it only operates at a statement level. I'm sure it'll be interesting though.

[1] https://github.com/vimeo/graph-explorer