I know PG 10 will have support "CREATE STATISTICS.." for this..
..but I wondered if there's a recommended workaround in earlier versions ?
Not without seeing the query....
2) memory explosion in hash join (due to poor estimate?) caused OOM.
As far as I know, the only way a hash join should do this is if the join includes a huge number of rows with exactly the same 32 bit hash codes. Otherwise, it should spill to disk without causing OOM. Hash aggregates, on the other hand, are a different matter.