Tomas Vondra <tomas.vondra@2ndquadrant.com> writes:
> Hmmm, I see the buildfarm is not happy about it - a couple of animals
> failed, but some succeeded. The failure seems like a simple difference
> in explain output, but it's not clear why would it happen (and I've
> ran the tests many times but never seen this failure).
Did you ever use force_parallel_mode = regress?
regards, tom lane