With the introduction of PostgreSQL 8.4, Bruce Momjian, a significant core developer, contributed a tool that can actually upgrade an entire database cluster in place. The time required is essentially only that necessary to copy the data files from the old installation to the new one. On a quick RAID system, this can be an order of magnitude faster than a dump/restore. The main drawback is similar to Slony: disk space must effectively be doubled for this upgrade method.
Postgresql hates itself. I’m convinced of this, and have considered the idea frequently over the years. I roll it around in my mouth just to savor the taste, only to hope the flavor changes eventually. A couple things have advanced, though not quite what one might hope.