"David G. Johnston" <david(dot)g(dot)johnston(at)gmail(dot)com> writes:
> A query you can execute should be incapable of failing if EXPLAIN for the
> same query is issued.
Indeed. This should be especially true for auto_explain, which isn't
even doing a re-parse or re-plan, but just dumping data from the
executor state tree for the just-finished query. Barring edge cases
like out-of-memory, it really shouldn't fail; so I see no reason why
we should consider major structural changes to make it (perhaps) less
likely to fail.
I continue to think that the most likely explanation is oracle_fdw
doing something it probably shouldn't be doing. I have no interest
in poking into that code myself, though.
regards, tom lane