"Scott Pederick" <sql(dot)postgresql(dot)org(at)pederick(dot)com> writes:
> Still can't fathom why it's not using an index scan on the jobs table..
Why exactly do you think that would make it faster?
The query evidently requires visiting every single jobs row, so a
seqscan seems appropriate to me; indeed I'd say the planner picked
exactly the perfect plan. If you think not, try forcing other plan
choices and see what happens to the runtime.
regards, tom lane