---
id: "lib-2026-10-09-the-reverts-will-continue-until-morale-improves"
type: "library"
title: "The reverts will continue until morale improves (PostgreSQL 19)"
lang: "EN"
org: "Peter Eisentraut"
source_type: "Practitioner post"
date: "2026-10-09"
week: "2026-W42"
url: "https://nosov.ai/library/lib-2026-10-09-the-reverts-will-continue-until-morale-improves/"
original_url: "https://peter.eisentraut.org/blog/2026/10/09/the-reverts-will-continue-until-morale-improves"
tier: "B"
tier_name: "Useful, limited"
tier_reason: "First-hand reflection by PostgreSQL committer Peter Eisentraut on why an unusual number of features were pulled from PostgreSQL 19; an informed insider view, explicitly his opinion, with few numbers."
summary_basis: "full-text"
original_minutes: "6"
tags: ["code review", "verification capacity", "open source"]
frames: ["verification-capacity"]
---

# The reverts will continue until morale improves (PostgreSQL 19)

Tier B (Useful, limited): First-hand reflection by PostgreSQL committer Peter Eisentraut on why an unusual number of features were pulled from PostgreSQL 19; an informed insider view, explicitly his opinion, with few numbers.

Summary based on the full text.

## In short

A blog post (9 October 2026) by Peter Eisentraut, a long-time PostgreSQL developer and committer of some reverted work. About 8 to 10 significant features were pulled from PostgreSQL 19 during beta, against a norm of about one per cycle. He gives several causes. One stands out: LLM-assisted code review now finds edge-case defects quickly, so features written years ago met a quality bar that rose after they were committed, and there was no time to clear the backlog of small fixes.

## Key numbers

- about 8 to 10: significant features reverted from the PostgreSQL 19 branch after beta began
- maybe one per cycle: number of reverts normally expected in a release cycle
- like forty: defect reports a single feature can face after LLM-assisted review
- half as many: major features PostgreSQL 19 now has compared with earlier expectations

## What it reports

- Some reverted features had serious architectural defects; many others had a long tail of minor edge-case bugs that would previously have been fixed gradually over about five years.
- LLM-assisted review can produce a list of around forty defect reports for one feature; even if each needs a three-line fix, the fixed overhead per report runs out the clock.
- The affected code predates useful LLM review, so he says this is not "vibe coding".
- The "Vulnpocalypse" (a wave of security work from April to August, the start of feature freeze to the latest security release) pulled senior developers off beta review, so many issues surfaced late.
- The community chose to keep the release schedule rather than delay for fixes.
- PostgreSQL's "everything works with everything" design creates combinations no one can enumerate by hand; he expects LLM-assisted fuzzing to become part of feature development and calls for sturdier internal interfaces and an "experimental" feature status.
- PostgreSQL 19 keeps REPACK CONCURRENTLY, logical replication of sequences and planner hinting with pg_plan_advice.

## How they got it

Personal reflection from inside the development process. No data on defect counts per feature or on which tools found them.

## Limits

- Opinion from one affected developer; he says the community has not finished reflecting.
- No breakdown of how many reverts were driven by LLM-found bugs versus architecture or staffing.
- No names of the review tools or measures of their false-positive rates.

## Why it matters for companies

Cheap automated review changes what "ready" means: code that passed human review years ago can suddenly carry dozens of known defects. Release plans need capacity for triage and small fixes when AI review is introduced, or the result is late reverts rather than better software. The same applies to internal codebases and to contracts and policies reviewed by AI. Teams should run AI review early in development, when fixes are cheap, rather than at release gates.

Original: https://peter.eisentraut.org/blog/2026/10/09/the-reverts-will-continue-until-morale-improves