[yaacovcorcos]

What Changed

Scient's browser now opens sites that refuse Electron, such as WhatsApp and monday.com, without a per-site list. This replaces the WhatsApp-only rule from #727.

apps/desktop/src/scient/browser/siteBrowserIdentity.ts has three rules, applied to every browser session, normal and incognito:

• Requests say plain Chrome in the User-Agent header. The page's own navigator.userAgent stays native.
• challenges.cloudflare.com keeps the true identity.
• A site that answers a page load with a 403 gets one reload with the true identity, remembered per host for that session.

This PR also carries upstream's 1280ccd0d2 (pingdotgg/t3code#28603), which pins the clock in one UsageService test. That test used a fixed date that passed the 92-day retention limit at 10:00 UTC on 2026-11-01, so it now fails on every branch, including main, and blocks the merge queue.

Why

Sites refuse opposite things, so neither identity works alone:

• WhatsApp refuses the app-name token and monday.com refuses Electron/.
• Full-page Cloudflare challenges and ScienceDirect refuse a Chrome header from a browser that isn't Chrome.
• Changing the identity for page JavaScript as well loops Cloudflare Turnstile, which is why upstream removed the old rewrite in fix(desktop): keep the native preview User-Agent so Turnstile passes pingdotgg/t3code#7110.

A site list only learns after a user is blocked. Changing only the header, with one fallback, covers both groups.

The retry reloads after the refused page arrives. Cancelling the response and calling loadURL shows ERR_BLOCKED_BY_CLIENT on a real address-bar load, and reloading any sooner repeats the previous page. If the user leaves first, the page that arrives instead isn't reloaded.

Measured (Electron 44.10 / Chrome 152, macOS)

Through the real address bar in a dev app, these load: WhatsApp, monday.com, IEEE, ScienceDirect, Science, Wiley, ACS, Oxford, Cell, Taylor & Francis, Sage, Nature, the Cloudflare login, two Cloudflare bot challenges, Overleaf, GitHub, Google, Microsoft and chess.com. A page that refuses everything gets exactly two requests.

Identity sent	dash.cloudflare.com/login	Full-page challenge	ScienceDirect
Whole identity plain Chrome	Loops (16 loads, 15 errors)	Not tested	Not tested
Plain header only	Fine	Refused	403
This change	Fine	Fine	Fine

The retry was run in real Electron against a local server: a plain refused page, one with a fragment, an empty 403, a redirect to a refused page, a page refused with both identities, and leaving or stopping before arrival.

The hooks cost about 0.02 ms per request, with no visible effect on a 300 MB download or WebSocket stream.

Not proven

• Sign-in flows other than WhatsApp's QR link, passkeys and video calls.
• Windows and Linux.
• ScienceDirect's own retry on this exact reload code: it didn't refuse on the test day, so the retry ran on the local server and in an earlier harness.

Linking WhatsApp by QR worked but took a long time. The busy process was the page's renderer, and the cause is unknown; the hooks aren't it.

On a challenged site's first visit, Back takes two presses because the challenge leaves its own history entry.

Electron keeps one listener per webRequest event per session, so a later listener on these sessions would silently replace this one. Nothing else registers one today.

Documentation impact: None — internal browser behaviour with no documented setting.

Checklist

• This PR is small and focused
• I explained what changed and why

🤖 Generated with Claude Code

[github-actions]

Thread transfer impact

✅ Thread transfer remains within every enforced ceiling.

Provider	Metric	Main baseline	This PR	Impact	PR ceiling
Codex	Total thread wire	5.5 KiB	5.5 KiB	0 B (0.0%)	6.8 KiB	✅
Codex	Thread snapshot wire	4.2 KiB	4.2 KiB	0 B (0.0%)	4.9 KiB	✅
Codex	Live turn WebSocket wire	1.3 KiB	1.3 KiB	0 B (0.0%)	2.0 KiB	✅
Codex	Live turn WebSocket decoded	22.4 KiB	22.4 KiB	0 B (0.0%)	24.4 KiB	✅
Codex	Live turn messages	2	2	0 (0.0%)	8	✅
Claude	Total thread wire	5.5 KiB	5.5 KiB	0 B (0.0%)	6.8 KiB	✅
Claude	Thread snapshot wire	4.2 KiB	4.2 KiB	0 B (0.0%)	4.9 KiB	✅
Claude	Live turn WebSocket wire	1.3 KiB	1.3 KiB	0 B (0.0%)	2.0 KiB	✅
Claude	Live turn WebSocket decoded	22.7 KiB	22.7 KiB	0 B (0.0%)	24.4 KiB	✅
Claude	Live turn messages	2	2	0 (0.0%)	8	✅

Baseline: 62ab2a9 · PR result: b52e405 · Source CI: success

Scenario and decoded snapshot size

10 historical turns, 5 command tools per turn, 878.9 KiB retained MCP result per historical turn, and a 1.05 MiB retained result in the measured turn.
Bounded compact HTTP snapshot with live cursor.

• Codex decoded thread snapshot: 88.2 KiB
• Claude decoded thread snapshot: 88.4 KiB

Updated in place by a trusted workflow. PR artifacts are strictly validated and never executed.