model-writtenrun 7549 - may be wrong, links go to the source messages. Commit-by-commit LLM analysis is in the Commits dialog.
- Aditya Desai reported a single-row lookup taking about 200 ms after migration from Oracle in #1; Justin Pryzby identified force_parallel_mode as a likely cause and suggested a testing-oriented name in #5.
- After Aditya confirmed the setting was enabled in #8, Bruce Momjian recommended disabling it and reviewing other configuration changes in #9; Aditya reported execution falling to 1 ms in #21.
- Tom Lane questioned why force_parallel_mode was classified as a planner option in #16, and Bruce supported moving it to developer options in #18, while Justin initially distinguished database-function debugging from PostgreSQL source development in #17.
- Justin submitted four patches in #25 to remove force_parallel_mode from the sample configuration and align several GUC categories, then removed accidentally included debugging changes in #27.
- Michael Paquier examined the patches in #28, supported the replication and autovacuum logging corrections, questioned the statistics and developer classifications, and requested an explicit autovacuum explanation after moving its logging setting; Justin defended removing force_parallel_mode from the sample configuration in #29.
- Michael committed the autovacuum logging portion of #27 as b094063cd1, with a clearer sample comment and adjusted placement, and announced the split landing in #31.
- Michael reconsidered renaming force_parallel_mode and broadening the Developer Options description in #31; Tom rejected renaming because it would break buildfarm configurations in #33, leaving documentation and classification changes as the chosen approach.
- Following Michael's wording proposal in #34, Justin posted the three remaining patches with a revised introduction in #35, and Tom supplied a shorter developer-testing description in #36.
- Michael committed the force_parallel_mode portion as ac725ee0f9, incorporating Tom's wording and adding the missing GUC_NOT_IN_SAMPLE flag, as explained in #38.
- Justin requested backpatching in #39; Michael considered the problem insufficient to justify it in #40, and Justin supplied earlier reports in #41, but no backport is recorded.
- Justin revisited the outstanding track_commit_timestamp classification and suggested removing the unused parent REPLICATION category in #42.
- Tom closed the thread in #43, reporting that the remaining category fixes had landed in a55a98477b through the later, broader cleanup in thread #70721; this thread's intended work was complete, with further questions continuing there.