OA — DNS Resolver (Fellows CodeSignal, Round 1)

First of two Fellows CodeSignal OAs (Spring 2026 rotation). 90 minutes, Python, proctored. Implement a simplified DNS resolver step-by-step against a fixed unit-test harness — early steps cover basic name lookup, later steps layer aliasing, fallback, error handling, cycle detection, caching, and concurrency. Requirements 90-minute proctored CodeSignal session (webcam + mic). Python only. The harness ships a partial codebase plus a multi-step spec; each step unlocks a unit-test group that must pass before moving on. The simulated DNS resolver must support, in roughly the order the steps add them: basic name normalization (case-folding, trailing-dot stripping) and direct A-record lookup recursive resolution through CNAME aliases fallback chains when the primary record path errors error propagation versus error swallowing (the spec is explicit about which step is which) cycle detection in alias chains response caching with a per-record TTL concurrent resolution — multiple in-flight queries against the same alias chain must coalesce rather than recompute Notes Difficulty is workmanship rather than algorithmic depth — closer to LeetCode Medium in raw skill but with the engineering surface area of a small library. Keep code structurally clean from step 1; otherwise later steps (cache, concurrency) wedge themselves into a brittle resolver. The proctor permits standard library only; no external resolver, no networking. The same Fellows cohort also receives the round-2 Python debugging OA (a separate session of 60 minutes; see the Fellows inventory entry). Preparation Implement, from scratch in under 45 minutes, a recursive name resolver over a static {name: record} map that handles CNAME chains and MX-style record-type dispatch. Layer a lru_cache-style TTL cache on top; verify that expiry is read-time lazy, not write-time eager, so the test harness's clock-advance semantics work. Build a tiny cycle detector using a per-call "visited" set; then convert it to a shared coalescing structure (a dict of in-flight Futures keyed by name) for the concurrency step. Refresh threading.Lock, concurrent.futures, and functools.wraps — the concurrency step is more idiomatic-Python than systems-programming.