🤖 tests: isolate renderer test globals

[ThomasK33]

Summary

• Stabilizes local renderer unit-test runs by keeping test-only module mocks and DOM globals scoped to the suites that need them.

Background

• bun test src runs many renderer test files in one process locally, so file-scope mock.module() registrations and direct globalThis.window/document overwrites can leak into later suites.
• The repro surfaced as useChatTranscriptFullWidth not calling its real API context after useModelsFromSettings, events.test seeing document as undefined, chatCommands inheriting a remote window.location, and ImmersiveReviewView rendering a stubbed diff renderer from ProposePlanToolCall.

Implementation

• Moves useModelsFromSettings and ProposePlanToolCall module mocks behind setup/teardown helpers that restore the real module exports after each test.
• Uses installDom() or explicit previous-global restoration so renderer tests leave the global DOM state as they found it.
• Makes events.test install its own DOM per test instead of relying on a one-time side-effect import.

Validation

• bun test src/browser/hooks/useModelsFromSettings.test.ts src/browser/hooks/useChatTranscriptFullWidth.test.tsx src/browser/utils/events.test.ts
• bun test src/browser/hooks/useRouting.test.ts src/browser/utils/chatCommands.test.ts
• bun test src/browser/features/Tools/ProposePlanToolCall.test.tsx src/browser/features/RightSidebar/CodeReview/ImmersiveReviewView.test.tsx
• bun test src/browser/hooks src/browser/utils
• make static-check

Risks

• Low product risk: this is test-only isolation. The main risk is that moving mocks into hooks could reveal tests that were unintentionally depending on cross-file pollution; the targeted cross-suite commands cover the known leak chains.

Pains

• make test now gets past the renderer failures but still exits on the pre-existing local Storybook budget guard (71 > 70 snapshot-enabled story files, 276 > 262 estimated snapshots), which is unrelated to this diff and not part of the CI unit shard.

---

Generated with mux • Model: openai:gpt-5.5 • Thinking: xhigh • Cost: $46.22

[ThomasK33]

@codex review

[chatgpt-codex-connector]

Codex Review: Didn't find any major issues. Breezy!

Reviewed commit: ccd7666ea3

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".