... no mention of pg_restore of any kind. Is there any reason why someone (maybe me) *shouldn't* rewrite this to include pg_restore?
I can't see any reason - it definitely should mention it.
Frankly, I think recommending psql to restore is a bad idea ...
Yes. And recommending pg_dumpall > sqlfile, but that goes hand in hand with that.
It also says that the least-downtime way is to use pg_dumpall in a pipe to psql. That's clearly not correct, since it does not support parallel restore (or parallel dump).
In short, +1 for you to write a patch that changes that.
It could probably deserve a better descirption of pg_upgrade as well, and an outline of the differences. Right now we spend the majority of the page on pg_dump, and then just say "oh, with pg_upgrade it only takes minutes"...
