Incremental Analysis: Round 8 (2026-05-14)
Summary
A single brief message from Tomas Vondra referencing an external thread about updating join_collapse_limit / geqo_threshold defaults. No new patch, no new benchmark data, no substantive technical argument beyond what was already covered.
What's New
Tomas references a separate thread [1] where Tom Lane asked whether it's time to update the join_collapse_limit/geqo_threshold defaults, and Andrei Lepikhov reported that join_collapse_limit=40 / geqo_threshold=14 worked acceptably for his case. Tomas observes:
-
Default updates may reduce GEQO trigger frequency: If
join_collapse_limitis raised (from current default 8) and/orgeqo_thresholdis adjusted (from current default 12), fewer queries would fall into the GEQO/GOO regime in practice. -
Implication for GOO's relevance: If DP becomes feasible for more queries via raised defaults, the population of queries that need a heuristic shrinks — though it doesn't disappear for truly large joins (20+ tables).
This is a contextual observation rather than a new technical argument. The message ends with "Agreed" — acknowledging a prior point without elaboration.