Playwright e2e specs are stale (24/26 failures on dev container)

[rubenvdlinde]

Baseline state (2026-05-16)

Full Playwright run against the local dev container reports 24 failed, 1 passed, 21 skipped out of 49 specs. Auth was fixed by #78 (globalSetup writes a shared storageState file, removed the global `OCS-APIRequest` header that broke browser-form login). Specs now reliably reach the app, but each one then fails on selector / content / route assertions.

Categories of breakage observed

1. Wrong landing path — many specs `page.goto('/apps/openbuilt')` which redirects to the Dashboard widget page (route `/`). The Virtual apps list (`.ob-app-card`) lives at `/applications`. Fixed in fix(#75): namespace manifest config.schema + test backfill + Playwright env #78 for `applicationCard` and `iconUpload`. Others still need updating.
2. Stale seed expectation — specs expect `hello-world` Application to exist with specific manifest content; sometimes the seeded data differs (esp. after versioned-model migration in [8/8] fix(tests): stabilise 11 PHPUnit failures + ProductionVersionGuardListener runtime bug #65).
3. Lifecycle-transition flow — `version-rollback` and `promoteDestructive` exercise publish/rollback paths that may have changed under ADR-002 (versioned-app model).
4. Component selectors — `page-designer` and `schema-designer` look for CSS classes that may have been refactored under the page-editor v1.1 work (openbuilt-page-editor v1.1 follow-up: stubs + validator marks + undo/redo + tests #9).

Suggested rollout

• One PR per spec category (icon, applicationCard, builderHost, ...).
• Each PR updates the spec + verifies on the dev container.
• Track separately from feature work because some specs assert behaviour that needs feature work in openbuilt-page-editor v1.1 follow-up: stubs + validator marks + undo/redo + tests #9 / Bootstrap follow-up: deferred verification tasks (4.3, 4.4, 5.2) #10 first.

Filed as a single tracker so the e2e debt doesn't get lost under the bug-fix queue.

[rubenvdlinde]

Resolved in #81 (merged). globalSetup + clean URLs + auth cleanup brought Playwright from 24/26 fail to 8/24 pass. The remaining failures are content-level / seed-data per-spec issues not infra; they'll be tracked per-test as they surface.