Step 4 — Principal Engineer Review

1. A team reports "our app is slow because the DOM is too large — we need virtualization everywhere." You have the containment measurement. How do you respond, what do you ask them to try first, and when would you agree with them?

2. Your design system ships defaults with !important "so they aren't accidentally overridden." Explain the consequence, and propose what to do instead. What is the migration path for consumers already working around it?

3. Design the cascade layer stack for a codebase with: a third-party widget you cannot modify, 5,000 lines of unmigrated legacy CSS, a design system, and application code. Where does each sit, and why is the legacy CSS's position actually convenient?

4. An engineer adds contain: strict to every component in the design system. What breaks? What would you require before accepting containment into shared components?

5. z-index: 9999 stopped working after a performance change. Explain the likely mechanism and how you would confirm it in under two minutes.

6. Your organisation chose runtime CSS-in-JS in 2019 because "our CSS was unmanageable at scale." Cascade layers now exist. Is that decision worth revisiting? Give the criteria, the cost of being wrong in each direction, and how you would run the evaluation without it becoming a year-long debate.

7. Theme switching causes a visible full-page flash. Explain the mechanism in terms of custom property inheritance, and give two fixes with different trade-offs.

8. Write the lint rules you would enforce on CSS in a large codebase. For each: what it catches, its false-positive rate, and why it is worth the friction. Include at least one that most teams do not have.

9. A token rename broke three applications at runtime with no build-time error. Design the token pipeline that would have caught it, and say what it costs.

10. Argue against this module's emphasis. When is CSS architecture not where the leverage is, and what would tell you that from a profile?

Answers in RESULTS.md. Questions 2, 3, 6 and 9 are natural ADRs for ../fe-00-roadmap/docs/decisions/.