test: remove low-signal tests

[flamboh]

Note

🤖 GPT-5.6 Sol on behalf of Oliver

problem

The test suite had grown to 133 executable files and more than 28,000 lines. Much of it asserted source snippets, Tailwind classes, pixel geometry, exact copy, or the same state at several layers.

solution

• delete 25 low-signal suites and prune 42 more
• remove all raw-source assertions and CSS or geometry checks
• consolidate repeated analytics, hydration, queue, share, logging, retry, Cobalt, manifest, and Tagium Save matrices
• keep binary format correctness, security, persistence, concurrency, privacy, destructive-action, cancellation, resource-lifecycle, and data-loss coverage
• update retained browser tests to use the current track action menu

The result removes 7,697 net test lines and 230 top-level declarations. Production code is unchanged.

human review

1. Run bun install, then bun run dev.
2. Import an MP3, edit its title, open settings, and return to the editor. The edited track and selection should remain.
3. Open a track action menu and remove the track. The confirmation should cancel safely, then remove the track when confirmed.
4. Start a SoundCloud set import with several tracks and remove one active and one queued track. Progress should update and removed tracks should not download.
5. Open bulk export confirmation. Focus should enter the dialog, Escape should dismiss it, and focus should return to the trigger.
6. In Tagium Save, switch themes, complete enough downloads to exceed five recent items, and try a picker response with audio. Theme persistence, the five-item limit, and picker audio should still work.

Worth checking manually: responsive layout, animation timing, copy, and visual truncation. This PR intentionally stops locking those details to exact classes or pixels.

Reviewer decisions: performance benchmarks were removed from the required Playwright suite, and repeated permutations now use representative boundary cases. Binary parsing, stream cleanup, authorization, signed URLs, manifest lifecycle, stale async writes, destructive behavior, cancellation, and active-resource cleanup remain covered.

verification

• bun run test: 100 files, 659 tests passed
• bun run lint: passed with existing warnings
• bun run typecheck: passed
• bun run build: passed
• Chromium Playwright suite: 23 tests passed
• Playwright discovery: 69 tests across Chromium, Firefox, and WebKit
• Firefox and WebKit were not run locally because their Playwright binaries are not installed

Changes made by GPT-5.6 Sol in T3 Code's Codex harness, with GPT-5.6 Luna subagents handling non-overlapping cleanup areas.

[github-actions]

React Doctor found no new issues. 🎉

_{Reviewed by React Doctor for commit 6f9b6d7.}

[cloudflare-workers-and-pages]

Deploying with    Cloudflare Workers

The latest updates on your project. Learn more about integrating Git with Workers.

Status	Name	Latest Commit	Preview URL	Updated (UTC)
✅ Deployment successful! View logs	tagium	6f9b6d7	Commit Preview URL Branch Preview URL	Aug 30 2026, 05:08 PM