Queries are based on site only, not the individual node at a site.

This speeds up queries and gives a fully-connected set more readily,
but there is no guarantee the nodes returned actually have data specific
to their node name (only site).