* Old implementation needs to see exactly the same rooms in exactly
the same order every time it loads new data, otherwise it will
create a duplicate entry for the room that differs only in display
order.
* New implementation ignores the display order when checking to see
if the room already exists.
