"Nick Fankhauser" <nickf(at)ontko(dot)com> writes:
> This apparently wasn't the source of problem since he did an analyze anyway,
> but my impression was that a fresh brand new database is exactly the
> situation where an analyze is needed- ie: a batch of data has just been
> loaded and stats haven't been collected yet.
Indeed. But as someone else already pointed out, a seqscan is the only
reasonable plan for an unqualified "SELECT whatever FROM table" query;
lack of stats wouldn't deter the planner from arriving at that
conclusion.
My guess is that the OP is failing to account for some client-side
inefficiency in absorbing a large query result.
regards, tom lane