Daniel Grace <dgrace(at)wingsnw(dot)com> writes:
> So I've spent the greater portion of the last two hours trying to slim
> down the schema and query enough to provide something that can
> reproduce this.
> While I can reproduce it 100% of the time with live data, I can't get
> it to reproduce at all with test data -- though I've included a few
> schemas below. It seems to be based on what plan ends up being
> constructed for the query.
Maybe you could force it for the test data with some suitable
combination of the enable_plantype switches.
regards, tom lane