Senior

Common interview patterns

• Move work off the critical path with `defer`, dynamic imports, and route splitting
• Prioritize hero images with `fetchpriority` and preload
• Cache at every layer: CDN, service worker, HTTP, in-memory
• Pick SSR/SSG/CSR per route based on freshness, personalization, and SEO

Common mistakes

• Optimizing TTI without checking what's blocking it (often a third-party script)
• Lazy-loading above-the-fold content and tanking LCP
• Treating bundle size as the only perf metric (TBT and INP often matter more)
• Preloading more than the browser can use, starving real-priority requests

Common interview patterns

• Split routes and big libraries with dynamic `import()`
• Use tree shaking + side-effect-free modules to drop unused exports
• Publish ESM + CJS + types for library consumers
• Generate stable source maps and ship only `.map` to error monitoring

Common mistakes

• Importing a whole library (`lodash`) when a named export would tree-shake
• Marking a side-effectful module as side-effect-free and losing CSS imports
• Shipping unminified bundles to production behind a misconfigured `NODE_ENV`
• Writing a webpack config that works in dev but breaks long-term caching in prod

Common interview patterns

• Pick monorepo vs polyrepo based on dependency velocity and release cadence
• Use module federation or build-time integration for micro-frontends, not iframes by default
• Version a design system with semver + changesets, ship migration codemods
• Expose an SDK with a stable, namespaced public API and internal-only escape hatches

Common mistakes

• Reaching for micro-frontends to solve a team-coordination problem that monorepo + codeowners would fix cheaper
• Publishing a design system without documenting the deprecation policy
• Letting internal-only utilities leak into a public SDK surface
• Coupling apps through shared global state instead of contracts