How would you implement a queue efficiently, and what's the naive mistake?

Use a ring buffer: a fixed array with head and tail indices that wrap, giving O(1) enqueue at the back and O(1) dequeue from the front with no element ever moving. The naive mistake is implementing the queue as a plain array where you enqueue at the end and dequeue by removing a[0] and shifting every remaining element down one, that makes each dequeue O(n), which is catastrophic in a hot path like a UART handler dequeuing per byte. The shift is pure waste: the data doesn't need to move, only your notion of "where the front is." The ring buffer captures that by advancing the tail index instead of moving data. So: a queue is a ring buffer; never a shifting array.