The median computation question serves as a practical interview filter that reveals how candidates handle API design, edge cases, and statistical reasoning without resorting to puzzle tricks. The author prefers it over brainteasers because the problem is straightforward — sort an array, find the middle — yet opens multiple technical avenues. Candidates must decide whether the function sorts in place or returns a new array, exposing their grasp of mutation semantics and performance trade-offs. The even-versus-odd branch creates a natural off-by-one trap, letting interviewers observe debugging style rather than just correctness.

Beyond syntax, the question invites discussion about why median often outperforms mean for skewed distributions, a useful signal for data-heavy roles. The provided Python reference implementation demonstrates clean error handling for empty inputs and leverages the standard library's sorting. Because the solution is readily testable, candidates can write unit cases on the spot, showing TDD habits. The exercise also surfaces standard-library awareness — `statistics.median` exists — letting senior engineers demonstrate pragmatism over reinvention.

This approach works because it scales with experience: juniors produce working code, mid-level engineers discuss complexity and mutability, and seniors connect it to production concerns like streaming medians or approximate algorithms. The question's strength lies in its simplicity — it's a FizzBuzz variant that doesn't feel like one.