WordPress 5.5 Absorbs the Best-Practice Checklist

Three plugins died of success this month, and we mark their passing with the respect the pattern deserves. WordPress 5.5 ships with XML sitemaps in core: every install now publishes a machine-readable content index by default. It ships native image lazy-loading: the loading attribute applied to images by default, so the browser defers off-screen media on its own. And it ships plugin and theme auto-updates, opt-in, per plugin, from the admin screen.

Each of these has been a staple of our deployment checklist for years. Until this month, each shipped as a plugin, a code snippet, or a maintenance-visit habit. Each is now simply what the platform does.

The individual features earn their field notes below. The release’s real story earns the finish, because it is the same story we told at Site Health last year, now with three more exhibits. The platforms keep absorbing the best-practice checklist. The baseline keeps rising beneath every install, regardless of who tends it. And the question worth an annual essay is what remains above the waterline when the absorption keeps coming.

Deployment notes from the field

Core sitemaps: serviceable, and mostly superseded on arrival. The native version covers the essentials (posts, pages, taxonomies, users) properly indexed and referenced. That makes it a real gift to the long tail of unattended installs that never configured an SEO plugin. For the professionally tended sites in this audience, the field note is anticlimactic in the correct way. The major SEO plugins’ sitemaps remain richer, with finer content-type control and the exclusion detail real sites need. They politely disable the core version on activation. The correct action for most maintained sites: none. Verify which sitemap is live, confirm Search Console points at it, and move on.

Native lazy-loading: the freest performance win in years, with one discipline attached. Browser-level deferral of off-screen images. No JavaScript library, no build step. It is already reducing initial page weight measurably on the image-heavy templates we have tested. The long-scrolling program pages and galleries benefit most. The discipline: width and height attributes matter more than ever. Lazy-loaded images without reserved dimensions are CLS generators, and the pairing of this feature with that metric in the same season is not coincidence. The platform is shipping the fix alongside the scoreboard. Audit templates for dimensioned media and the feature pays immediately. Skip the audit and it pays partially while charging your CLS.

Auto-updates: the feature that deserves actual thought. It gets its own section, because reflexive enthusiasm and reflexive refusal are both wrong and both circulating.

The auto-update policy question, reconciled

• The tension, honestly: the post-Drupalgeddon religion says patch speed is survival, written in exploit timelines. The change-control discipline of enterprise sites says untested changes to production are how outages happen, equally earned. Auto-updates sit exactly on the fault line, and a blanket policy in either direction ignores one truth to honor the other.

• The grown-up answer is risk-tiering: auto-update the low-risk tier, the single-purpose utilities whose failure modes are cosmetic. Gate the critical tier behind the staging-and-scheduled-window path. The commerce stack, the membership system, the page builder: anything whose failure is an incident. Write the tiering down as policy, per install, in the maintenance docs.

• The unattended tier changes the math: for the long tail the census always finds (the department site, the archived-but-live property) turning auto-updates on is flatly better than the alternative. The alternative is no updates ever. The WannaCry casualty record argues for setting the orphans to automatic and treating any breakage as cheaper than a hack.

• Backups and monitoring are the enabling conditions: auto-anything without restore-tested backups and uptime checks is gambling with extra steps. The policy rests on the same continuity items we keep demanding, and 5.5’s feature list does not include them. That is the point the finish will generalize.

The pattern, annual essay edition

Now the release’s actual significance, continuing what is apparently an annual summer series. The trend line is clear and speeding up. Performance and security defaults keep migrating into the platforms themselves. Each migration shrinks the gap between the well-maintained site and the neglected one, for the fraction of the checklist that software can own.

The absorbed tier keeps growing. The sitemaps and lazy-loading of this release. The health dashboard and recovery mode of last year’s. The HTTPS provisioning the hosts folded in. The caching the managed tiers now include. And the same motion running at enterprise scale, where the platform now enforces what consultants once begged for.

Our position on the trend: unreservedly in favor, including of what it does to our own industry’s lower tiers. Work that reduces to a checklist should become a default. Practices that priced themselves on performing checklists are being correctly automated out of their invoices. The professional tier that remains is the one the absorption cannot reach: the governance fraction. Whether anyone is watching. Whether backups restore. Whether the critical plugins have maintainers. Whether ownership is named, and whether the policy exists and is followed. The platform can nag, default, and enforce. It still cannot supply the human commitment that turns capability into reliability. It always ends there. It ended there last year, it ends there this year, and the releases keep obligingly clearing away everything else so the residue is impossible to mistake.

The practical summary

For the maintenance calendar, this quarter: verify sitemap arrangements and Search Console. Audit image dimensions before trusting the lazy-loading gain. Write the auto-update tiering policy per site, defaulting the orphans to on. Confirm the backup-and-monitoring preconditions that make any of it safe.

Each of these annual posts now ends the same way on purpose, because the repetition is the finding. The software keeps getting better at being operated. The gap between tended and untended keeps narrowing at the layers software can reach. The differentiating layer keeps rising toward pure governance, which no release will ever ship.

Three plugins died of success this month. The checklist they served is shorter. What remains on it was always the part that mattered, and it is now, conveniently, most of what is left.