> Are there tests that offer some specific examples indicating when a conformed table is used versus a source?
> Does the planner consider only the query structure when deciding to use a conformed table?
Yes it's based upon pushdown and it's more heuristic than a full search problem. That is as we attempt to push joins down we'll opportunistically check to see if the join can be pushed to the nominal target or if there is a common subset of sources considering the conformed sources as well. We revert back to normal planning once the set of possible targets is back down to 1.
Are there tests that offer some specific examples indicating when a conformed table is used versus a source?
Does the planner consider only the query structure when deciding to use a conformed table?