I ran some experiments, and with the queries that I was
testing with, the clustered rtree was about 30% faster than
the unclustered one.
> This isn't really relevant to your main point, but: since an rtree
> doesn't have an associated sort order, it's not clear to me that this
> operation makes any sense. Have you determined that you'll actually get
> any performance improvement as a result of the clustering? I suspect
> you may find that you're just rearranging the table into a different
> random order.