[risu729]

Popup cases now live in four passive modules, reducing the discovered entry from 2,356 to 335 lines. Seven synchronous registrar calls preserve the 85 cases, root cleanup, mock/fixture ownership, CSS/import order and initializer positions. The sole late lineText binding remains a callback-only lazy getter. All SSO cases are retained; this PR depends on SSO coordination. Only this entry's line-limit exemption is removed, with ownership documented.

Independent source review is clear at final HEAD. The reviewer independently reconstructed all seven registrars and matched the complete TypeScript/JSX AST, 85 tests, 149 assertion sites, two JSX elements, raw literals and 94 comments. Every capture/import owner/type classification and initialization order was checked. The helper modules have no executable initialization or runtime entry cycle. Options typecheck and scoped static/docs/SDK-source/secret checks pass.

After the actual cloud browser attempt, separate verification used repository-pinned Chromium with synthetic clients. The original Git-base entry was injected before JSX transformation at the identical Vite module path: 85/85 pass in 25.70s. The actual existing options task with coverage passed 460/460 across 26 files in 151.75s (174.10s with its probe/API build dependencies). All 85 IDs, full title paths, order and outcomes match, with no added suites. Complete beforeEach/afterEach start/end sequences match, including 85 completed root cleanups. Discovery remains 26 entries; coverage contains 305 production files and no helper modules. Two initial baseline setup attempts stopped before cases were collected (missing generated WXT types, then initial Vite optimizer reload); existing preparation and unchanged ordinary cache resolved them. The successful full browser task ran once.

Final CI passes Verify, Orders and components, all six packaged-browser shards and CI Check; CodeQL passes. The actual tested merge has the exact SSO base and final head as parents and its tree equals final HEAD. Compared with exact-base CI, every task count matches: Verify 10,212/327 plus 73 tools; components 1,981/119. Observed Verify 273→191s, runtime 225→158.21s, components 247→221s, options 142.31→127.02s and entry 18.703→11.872s. These single observations do not establish causal speedup or zero-regression claims.

No production, task, discovery, coverage, timeout or expectation changes. No user-browser, real-account, paid-provider, deployment, merge or auto-merge operations.

[risu729]

Independent source review: CLEAR at final HEAD.

Independently reconstructed all seven registrars without author proof metadata. The complete TypeScript/JSX AST matches the baseline after restoring the single context.lineText access to its original lexical reference. All 85 tests, 149 assertions, two JSX elements and root cleanup are preserved. Every capture was checked: fixture/helper identities, initializer positions and registration order match. The exact callback-only lazy getter preserves the sole late binding. Raw literals and all 94 comments are preserved. Import owners and type/value classification match; modules have no executable initialization or runtime import cycle. Discovery, coverage, tasks, dependencies and production code remain unchanged. All five files meet the line limit; owning documentation is accurate. The SSO base remains an ancestor and the final checkout is clean.

Static source review only. Separate browser runtime comparison and full CI are tracked independently.