The WordPress Performance Profiling Toolkit We Actually Use
Not a plugin roundup. Four tools that actually found something on a real WordPress VIP network, what each one is blind to, and which symptom sends you to which one.
performance
What ‘VIP-Style Multisite’ Actually Means for a Performance Bug
Every fix in this series landed at a different layer — application code, shared plugin, edge config, vendor dashboard. The scoping decision matters more than the fix itself, and it’s rarely obvious from the bug report.
Sizing a Covering Index on Shared Multisite MySQL
An index that fixes one site’s slow query has a write cost paid by every site sharing that database tier. A checklist for deciding whether a covering index actually pays for itself, before adding it.
Why Your Full-Page Cache Needs an Opinion About Query Strings
Two separate incidents on the same network turned out to be the same root cause wearing different clothes: a cache key that let the request decide what it means, instead of the other way around.
Shipping a Database Index Without an Outage on WordPress VIP
A shared plugin update needed five new indexes across a multisite network. The bug that almost shipped wasn’t in the indexes — it was in assuming a diff between two versions was the same thing as the full version.
Profiling What Googlebot Actually Sees: A BigQuery Log Deep Dive
Search Console said one site’s average crawl response time was roughly double a near-identical sibling’s. The access logs disagreed — until they were split by bot type, and a cache-busting tracking parameter fell out.
Turning a Debugging Script Into a WP-CLI Profiler
The SAVEQUERIES mu-plugin from the last post left you with a raw log file and a promise to wrap WP-CLI commands around it. Here’s the tool, not just the promise.
When Your Performance Plugin Causes the Regression It’s Supposed to Prevent
A plugin built to delay non-critical JavaScript until user interaction concentrated all of that JavaScript’s cost into exactly the moment interaction responsiveness gets measured.
The LCP Element That Wasn’t Content
Weeks of hero-image optimization went nowhere, because the thing Lighthouse was actually measuring wasn’t the hero image at all — it was a cookie banner’s logo.
One Bot Request Poisoned Our Cache — And Every Real Visitor Got the Broken Page
Ads were randomly missing on one site, with no reproducible pattern anyone could find. The cause was a single automated request with the wrong Accept header, cached and served to everyone after it.