Comment 4 for bug 1738488

The biggest issue with this flat_fielder query is that it's causing, like, 20 joins. That's well past both the join collapse limit and the GEQO threshold in any reasonable setup. Many of the joined tables seem duplicated, and I think we could probably trim those down significantly by only following one path to any particular table.

If we can get the list of joins down to just what we actually need, instead of providing separate paths to, for instance, bre, the obvious selectivity of the usr filter will probably be noticed by the planner. I tested this by pushing the join collapse limit and GEQO threshold way up to let the planner look for a better join order, and it did exactly that.