perf(renderer): unmount closed setup guide content

[nwparker]

Summary

• keep only a one-selector setup-guide shell mounted after the guide has been used
• retain the full content for the 200 ms close animation, then release its progress/runtime subscriptions
• preserve dismissed-close telemetry and cancel a pending unmount when the guide reopens

This follows the mount-gating pattern from #13525. The intentional always-on setup telemetry observer remains; this removes the second progress graph retained by the closed modal.

ELI5

The Getting Started guide is like a checklist teacher. After you opened it once, Orca kept a second copy of that teacher listening to project, remote-host, skill, and integration updates even after the guide disappeared.

Now the closed guide keeps only a tiny doorbell that knows whether it should open. The teacher stays for the brief closing animation, then goes home until the next open.

Visual explainer

flowchart LR
  U[Project or remote-host update]
  O[Intentional telemetry observer]
  M[Retained closed modal]
  P[20 direct progress selectors plus runtime hooks]
  N[Render nothing]
  V[One visibility selector]
  X[Content unmounted after 300 ms]

  subgraph Before
    U --> O
    U --> M --> P --> N
  end

  subgraph After
    U --> O
    U --> V --> X
  end

Loading

Improvement proof

• useSetupGuideProgress alone owns 20 direct Zustand selectors; the retained modal also owned modal/runtime subscriptions.
• The regression harness models a representative content selector and raw store subscription. On base, the closed component still mounts that content and the new test fails.
• After the 300 ms linger: active content subscriptions = 0.
• 1,000 closed store writes after unmount: 0 content renders and 0 content notifications.
• At 299 ms the content is still mounted for the close path; at 300 ms it is released.
• Reopening at 299 ms cancels the pending unmount and keeps exactly one content subscription.

Testing checklist

• SetupGuideModal.mount-gating.test.tsx: 2/2 passed
• focused setup-guide suite: 22/22 passed
• pnpm run typecheck: passed
• changed-code quality gate: 0 new findings across 2 files
• reliability manifest: 73/73 gates passed
• max-lines ratchet: no new bypasses
• git diff --check: passed

Full-suite check: 49,130 tests passed. Five unrelated files failed in the heavily parallel run. A serial rerun cleared the native-chat watcher and worktree-directory poller failures. Three environment/load failures remained:

• cross-version fixture extraction: missing generated .staging-v799-*/src
• terminal frame equivalence fuzz: existing 30 s timeout (44.9 s on this machine)
• GitHub project source-context import test: existing 30 s timeout

The changed setup-guide test was rerun after the full suite and remained green.

No-regression review

• Closing still gives telemetry an explicit open=false render, so it records dismissed, not interrupted.
• The 300 ms linger covers the dialog's 200 ms exit animation.
• Timer cleanup covers reopen and component unmount; the 299 ms reopen race is tested.
• Installed-skill discovery remounts from its runtime-scoped module cache, so this does not force a new scan.
• Existing setup-script and remote-runtime cancellation/stale guards remain unchanged.
• Active-step state already reset on reopen, so unmounting adds no user-visible state reset.
• No user-facing regressions found.

Screenshots

Not applicable: this changes hidden component lifetime only; visible layout and styling are unchanged. The lifecycle diagram above is the visual proof.

AI review report

• P0: none
• P1: none
• P2: none
• Checked failure cleanup, timer concurrency, stale async work, close telemetry, repeated opens, macOS/Linux/Windows behavior, SSH/runtime ownership, and folder-workspace assumptions.

Security audit

No authentication, authorization, persistence, network sink, shell execution, dependency, or user-data surface changed.

Notes

• No mobile-facing surface or remote wire contract changed.
• No Git/Git-provider behavior changed.
• Development StrictMode replays telemetry effects when a newly remounted subtree is probed; contributor telemetry transport is disabled, and production React does not replay effects.

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Repository UI

Review profile: CHILL

Plan: Pro Plus

Run ID: 851757d4-d909-4b83-89c4-c6785443cf01

📥 Commits

Reviewing files that changed from the base of the PR and between 1310102 and 53b9078.

📒 Files selected for processing (2)

• src/renderer/src/components/setup-guide/SetupGuideModal.mount-gating.test.tsx
• src/renderer/src/components/setup-guide/SetupGuideModal.tsx

---

📝 Walkthrough

Walkthrough

SetupGuideModal now keeps its content mounted for 300 ms after closing. The wrapper passes the current open state to the content component. Telemetry, step-selection effects, completed-step handling, and dialog visibility use this state. Tests cover delayed unmounting, suppressed closed-state updates, subscription cleanup, and reopening during the close linger.

🚥 Pre-merge checks | ✅ 5

✅ Passed checks (5 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title clearly identifies the main change: unmounting closed setup guide content to reduce retained subscriptions.
Description check	✅ Passed	The description covers the user-visible change, testing, AI review, security audit, platform checks, risks, and known environment failures.
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.

✨ Finishing Touches

📝 Generate docstrings

• Create stacked PR
• Commit on current branch

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}