fix(desktop): hold the launch logo until a surface commits ready content

[Astro-Han]

Summary

#5494 removed the pre-mount onboarding prefetch, so AppShell's first commit now lands before any data resolves. The .maka-preload logo dismissed on that commit (body:has(#root > *)), and the shell then rendered the .maka-onboarding-loading skeleton card while onboarding:getSnapshot waited behind the Host boot gate — two loading surfaces over the same gap, visible as logo → skeleton card → app.

Collapse them into one. The overlay now fades only when a surface commits data-maka-content-ready:

• .appFrame sets it once isOnboardingLoading clears — the exact boundary where the skeleton used to disappear (first snapshot settled, sessions found, or load-error fallback).
• WorkHub's composer sets it at mount (no equivalent data gap).
• The handoff attention dialog and the error surface set it immediately — they cannot wait on the data they gate.

With the overlay covering the gap, .maka-onboarding-loading is deleted along with its styles, prop, and smoke-diagnostic selector. The 8s wedged-bundle timeout is unchanged. notifyRendererReady's comments are corrected: its live job is the crash-recovery reload, not first paint.

Refs #5494

Verification

Real Electron launch, CDP screencast on a fresh profile — BEFORE: the logo fades on first commit and a bare skeleton card sits in the chat area until the snapshot lands; AFTER: the logo holds until the settled hero is committed, then fades. The .maka-onboarding-loading card never appears.

BEFORE — logo dismisses on first commit, bare skeleton card shows:

launch-before.mp4

AFTER — logo holds until the settled hero commits:

launch-after.mp4

DOM timeline sampled every ~80ms confirms the ordering: .appFrame mounts at ~300ms with the overlay still at opacity 1; data-maka-content-ready commits with the hero at ~980ms; the overlay is hidden by ~1.1s.

• build:workspace-deps, build (renderer + main + overlay), typecheck — pass
• node --test dist/main/__tests__ — 2770/2770, including a new contract test pinning the selector↔emitter pair and the skeleton's removal
• npm run format, npm run lint — clean
• Not run: full npm test, Playwright e2e (no native-boundary behavior changed)

AI use

Select exactly one:

• No generative tool made a substantive contribution
• Generative tooling made a substantive contribution

Tool(s) and scope: Devin — diagnosis, implementation, regression test, and launch recordings.

Checklist

• Tests cover the change and fail without it
• Lint, format, typecheck and the affected suites pass locally

Does this PR entail a change in behavior?

• Yes — described under Summary above
• No

[jackwener]

I reviewed 4024c13a770a0a802fd10ca480b1631471851f7d.

No P0–P2.

The launch overlay now fades only when a surface commits data-maka-content-ready. AppShell sets that once isOnboardingLoading clears; WorkHubRoot only mounts on the WorkHub surface, so it cannot dismiss the desktop logo early. The handoff attention dialog and the error surface mark ready immediately. The second skeleton (.maka-onboarding-loading) is gone. The 8s wedged-bundle timeout is unchanged.

Hosted test was still queued at this review. I approved on the code, not on that check.

简体中文

我审查了 4024c13a770a0a802fd10ca480b1631471851f7d。没有 P0–P2。

启动 overlay 等到表面写上 data-maka-content-ready 才褪。AppShell 在 onboarding 加载结束时写；WorkHub 只在自己的 surface 挂载，不会提前关掉桌面 logo。第二个 skeleton 删了。8 秒超时还在。本审查时 hosted test 仍在排队，是按代码批的。

---

Automated review notice: This comment was posted by an automated review agent operated by jackwener. It is not an independent human review and does not replace one.