Performance Chat Summary: 5 May 2026

The full chat log is available beginning here on Slack.

Table of Contents

Toggle

WordPress Performance Trac tickets

@spacedmonkey asked whether commits to trunk are currently allowed or if only RTC-related changes should be committed. @westonruter clarified that trunk is still frozen, except for 7.0-specific fixes that get back-ported, and testing commits.
- @spacedmonkey said they would hold off committing any performance-related changes for now and asked others to ping them if anything needs review or commit.
- @westonruter added that @pbearne already has a few PRs for testing changes that could be moved forward for commit now, though not performance-related.
- @pbearne mentioned that there are more PRs to come.
@westonruter shared that there is about a 10% regression in TTFB in WordPress 7.0 compared to 6.9, based on benchmarking noticed by @mukesh27. @westonruter mentioned that TTFB-LCP does not show a regression, so this appears to be additional PHP processing slowing things down, but no single cause has stood out yet.
@westonruter also shared ticket #65165, which was recently opened, about script modules depending on classic scripts. @westonruter noted that this is related to performance because it can reduce the amount of scripts loaded on the page thanks to dynamic imports, which are non-blocking, and mentioned that the ticket had just come in and had not yet been reviewed in depth.
@westonruter further pointed out ticket #64696, which focuses on improving the scalability of real-time collaboration via HTTP polling and its impact on persistent post caches, noting that things seem to have gone a bit quiet on that ticket.

Performance Lab Plugin (and other performance plugins)

@westonruter shared that PR #2461 which updates @wordpress/scripts and related packages while fixing backward compatibility issues, is currently top of mind.

Open Floor

@westonruter shared a LinkedIn post highlighting that appending tags late in the can be too late for optimal performance if an initial chunk of HTML is sent without those tags, which can happen when a lot of CSS is inlined. @westonruter noted that this is relevant for Optimization Detective, since it currently appends these preload tags to the end of the head.
- @westonruter mentioned that HTTP Link headers are also being sent, so in practice this might not be an issue. However, @westonruter pointed out issue #2304, where large Link headers can exceed Nginx limits and cause 502 errors.