WebSpeed Optimize — observed font delivery before/after, 25 October 2026

Target: https://webspeedoptimize.com/es/
Six real public loads per version, fresh Chromium 152.0.7977.82 per load,
1280x900 CSS viewport, device scale1, es-ES, reduced-motion preference.
CDP emulation:150ms latency,200000B/s download,93750B/s upload,CPU slowdown4.
Browser cache disabled. This does not guarantee cold DNS,OS,connection or CDN caches.
Fresh anonymous contexts, no consent choice and no page interaction during sampling.

Before: external Google Fonts stylesheet and Sora WOFF2.
After: same original Sora WOFF2 files served by the site, with font-face CSS inline.
Original fonts and SIL Open Font License retained. No preload added.
Original theme source and9nativepost/meta records remained unchanged during this
intervention. Main text and heading family/weight/size/geometry match across12loads.

LCP candidate and CLS session-window maximum were observed through an initial
approximately8second window. These are desktop lab observations, not physical-phone,
full-lifecycle field Core Web Vitals,INP,Lighthouse scores,Google clicks or sales.
All raw loads and response context are included; no run is silently discarded.
Before LCP-candidate median2396ms,range2300–2512;after2316ms,range2224–2608.
Ranges overlap, and the worst after load exceeds the worst before load. The80ms
median difference is not a universal guarantee or randomized causal estimate.

The two series ran sequentially:before08:53:10–08:54:06UTC and
after08:58:01–08:58:59UTC. Server/network variation remains possible.
Payload transfer and HTML quotation checks were corrected before/post activation;
no measurement sample was replaced to obtain a preferred result.

Files: samples.csv, evidence.json, README.txt. No credentials,session cookies,
executable code or customer data. Font binaries are not part of this package.
