Does anyone have any ideas? All data are loaded into this table via copy and no updates are done. Autovacuum settings weren't changed (and is on both). Do I need to increase shared_buffers to half of available memory for the planner to make certain optimisations? Anything else I'm missing or can try? The new server has been running for almost two weeks now so I would have thought things would have had a chance to settle down.
It is looking like some missing optimization that was removed from RC release.