Stop the site issuing five requests it knows will…

Details

Three URL faults, all of them requests that could never succeed. Measured over
`/`, `/v1/`, `/v2/`, `/changelog/`, a deep changelog page and `/view.html`:
**The explainer asked GitHub before the file next to it.** Its dataset sources
listed two `raw.githubusercontent.com` URLs ahead of the local copy, but
neither PyEDCR branch carries `examples/web/`, so both 404 on every dataset
load — four failed cross-origin round trips per view of `/` before the copy
committed with the site was ever tried. Local now leads; the repository stays
as a fallback. Verified in a browser across a dataset switch: 2 local fetches,
**0 requests to GitHub, 0 failed requests**.
**The document viewer declared no favicon**, so a browser fell back to probing
`/favicon.ico` and got a 404.
**Nothing was served at `/favicon.ico`.** Browsers and crawlers probe that path
on any origin whether or not a link tag is present, so every one of those was a
404 regardless of the four layouts that do declare one. A copy now sits at the
site root.
After: zero 4xx and zero failed requests across all six page types, and no
redirects anywhere. (Local builds still show `google-analytics.com/g/collect`
failing — that is the `G-PREVIEW0000` placeholder ID, not a site fault.)
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01DAJJ63FFnLoD5XLpu6yJaG

Files changed (3)