On 2018-02-27 16:58:11 -0500, Tom Lane wrote:
> > Also, I recall discussion that select statements will touch the visibility
> > map (hence causing write I/O even in a read-only query) but [1] indicates
> > that only vacuum will set them ddl will clear them.
>
> Hm, I don't recall that, but I've not been involved in the last few rounds
> of hacking on that mechanism.
I'm fairly certain that only vacuum and table rewrites like cluster sets
all-visible, and that there was never any released code that did so
during SELECT. I think there were a few patches debating whether we
could change that, but they never really got anywhere afair.
Greetings,
Andres Freund