perf: accelerate login and theme background loading

[bifrost0x]

Problem and outcome

Login backgrounds appeared noticeably after the rest of the page, and the first theme switch downloaded a 1.2-2.2 MB PNG. Serve full-resolution WebP backgrounds and start the selected authentication background immediately after resolving the saved theme, instead of waiting for window load and an idle callback.

The ten served images total 857 KB instead of 16.88 MB (94.9% less). Original PNGs are retained. WebP encoding is lossy at quality 90 (89 for Arctic), with unchanged 1672 x 941 dimensions.

Cold-cache Chromium measurements at 5 Mbit/s and 40 ms latency, median of three runs:

Background response completed	Before	After
Login, Glass, from navigation	3.172 s	0.483 s
Login, Paper, from navigation	3.780 s	0.774 s
First Glass-to-Paper switch	3.242 s	0.296 s

These measure background downloads, not whole-app responsiveness. The switch measurement isolates the CSS theme assignment; normal settings persistence is covered by browser tests.

Scope and links

• Change ten CSS asset URLs and content hashes; preserve layout, colors and effects.
• Keep the initial CSS guard so the saved theme is resolved before requesting an image.
• Register image/webp explicitly for hosts without a correct MIME mapping.
• Add image-size, decode, early-loading, storage-fallback and HTTP delivery regression coverage.
• No authentication, SSH/SFTP, terminal-input, permission, database, dependency or deployment configuration changes.

Validation

Head: 635be4228a5d8e865926ff653bf9ead45c34e466; base: c0f83a85c3e8ddd3114c528bb57f14c2a6eb764f.

• 677 JavaScript tests passed; ESLint, vendor integrity and diff checks passed.
• 20 static reference/delivery tests passed, including all ten WebP MIME types, bytes, immutable versions, ETag/HEAD, cookie isolation and a deliberately incorrect host MIME mapping.
• 11 focused Chromium tests passed, including decoding all backgrounds and usable login when storage and image loading fail. Screenshots inspected for Glass and Paper.
• Full Python 3.12 run: 3400 passed, 53 skipped, 6 failed. All six failures reproduced on the unchanged base: three test_entrypoint.py cases, the oversized-command shrink case, the Linux PTY directory probe and the Linux Tailscale route lookup. The additional host-MIME regression passed in the focused run.
• Full Chromium run: 183 passed, 1 failed. product-captures.spec.js expects the Settings content bottom within 1080 px; both unchanged base and candidate report 1141 px. The two additional asset/fallback tests passed separately.

Local suites are not entirely green due to these reproduced baseline failures. Cross-browser and live remote-server integration were not revalidated. Hosted checks remain authoritative for this head.

Delivery

Prepared as a draft for review. No merge, deployment or versioned release is included.

[chatgpt-codex-connector]

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-10-04T11:10:13.902608Z	635be42	Draft marked ready

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.