Design a cross-product feature flag strategy for coordinated release

WHAT THIS TESTS: This question tests whether you understand that feature flags are a distributed systems problem once they cross process boundaries. The interviewer cares about consistency models, dependency management, fault tolerance, and user-experience integrity across independently deployed services. They want to see that you think in terms of end-to-end user journeys rather than isolated code paths.

A GOOD ANSWER COVERS: First, namespace and dependency design. You should propose a unified flag registry or at least a naming convention that makes cross-product dependencies explicit, such as a prefix like productB.checkout_api_v2 that Product A imports as a prerequisite. Second, evaluation strategy. Recommend evaluating the flag once at the edge or in the entry service and propagating the decision through request context or a secure header so downstream services do not re-evaluate independently and create split-brain states. Third, configuration plumbing. Describe a central config store with local caching, TTLs, and circuit breakers so that a network partition does not default every request to on or off unpredictably. Fourth, rollout mechanics. Mention a coordinated ramp using user-scoped bucketing, canary segments, and kill switches that operate on the entire cross-product flow as a single unit.

COMMON WRONG ANSWERS: The biggest red flag is suggesting each product manages its own flag independently with no dependency tracking. This leads to partial rollouts where a user sees the new checkout in Product A but hits the old inventory API in Product B. Another mistake is proposing synchronous flag checks across services without caching or fallback logic, which adds latency and creates a single point of failure. Saying you will just hardcode a dependency version or deploy both products simultaneously as a fix also misses the point because it ignores independent deployability.

LIKELY FOLLOW-UPS: The interviewer may ask what happens if Product B rolls back while Product A keeps the flag on. They may also ask how you handle flag evaluation in mobile clients versus server-side, or how you audit and observe the state of a cross-product flag in production. Be ready to discuss idempotency keys, dark launches, and how you would measure error budgets during the coordinated rollout.

ONE CONCRETE EXAMPLE: Imagine a new checkout flow in Product A that requires a real-time reservation endpoint in Product B. You create a single gate checkout_v2_reservation that lives in a shared config service. Product A evaluates it when the user begins checkout and stamps the decision into a X-Feature-Context header. Product B reads that header and enforces the same branch without re-evaluating the flag. If Product B's new endpoint errors above one percent, a circuit breaker falls back to the legacy path and the header acts as a kill switch for both products simultaneously.