The release was once expected in late September or early October.
It is now expected to be generally available around November 19.
According to The Register, developers have dropped between eight and 10 significant features, which are now likely to move to PostgreSQL 20.
Peter Eisentraut, a vice president at the PostgreSQL support firm EDB and a leading PostgreSQL committer, said in a blog post that the reverts could be at least partly attributed to the LLM-assisted code reviews developers have been using.
LLM-assisted review also surfaced a long tail of mostly minor edge-case problems that, in the past, might not have been noticed until years after release.
He said that even if each needs only a small fix, every one carries fixed overhead, and a list of around forty defect reports can exhaust the available time.
He argued that the effective quality bar rose after the features were written and committed, because AI-supported review became more useful in the months before launch.
He also stressed that the problems were not caused by "vibe coding," saying that approach was not available when most of the features were written.
One feature pulled was support for SQL Property Graph Queries (SQL/PGQ), a part of the SQL standard since 2023 that provides syntax for exploring relationships between nodes linked by edges. Contributor Tom Lane had earlier warned that shipping it in version 19 would likely bring post-release bugs that could not be fixed until version 20. Eisentraut says PostgreSQL 19 will still be a great and very robust release, including REPACK CONCURRENTLY, which reclaims disk space without the table lock that VACUUM FULL requires. PostgreSQL 20 had been expected to contain only bug fixes, but it is now likely to carry the features held back from this release. Eisentraut expects the situation to ease in future cycles if LLM review capabilities keep improving at the current pace.