Description
I run a WooCommerce store on Convesio (WP Cloud / Automattic Atomic Platform, Sydney). The site has a well-defined performance problem that needs profiler-level diagnosis, not generic PageSpeed advice. The problem: Around 14 February 2026 the site’s p75 TTFB stepped from sub-1s to 5s overnight. It has since been partially remediated to ~3.7s but is still in the CrUX “Poor” range. Symptoms today: • Checkout POST (wc-ajax=checkout) consistently takes 9 to 10 seconds • Mobile PageSpeed LCP 11.6s (most of my traffic is mobile) • Homepage CWV field data: LCP 5.2s, TTFB 5.5s (fails) • Category pages pass CWV (well-cached edge layer) • Regression on 14 Feb has never been explained by the host Stack: • Hosting: Convesio (WP Cloud infrastructure, Sydney edge) • Theme: Divi • ~30 active plugins (Fluid Checkout, Fast Cart by Barn2, WooCommerce Product Bundles, Wordfence, Yoast, Omnisend, GTM4WP, Metorik, MyCryptoCheckout, WPS Hide Login, and others) • FlyingPress (asset optimisation only) • Cloudflare free tier (confirmed not in AU request path) What’s already been ruled out: • Convesio APM confirmed 95% of execution time is PHP within WordPress, not database, not server, not external APIs • Sold Out Badge for WooCommerce removed (was firing 395 queries per page) • Divi performance settings reconciled with FlyingPress to avoid double-optimisation • Cache layer architecture mapped (Atomic edge / Batcache / FlyingPress) • Direct A/B header testing confirmed FlyingPress is not the source o