Loved the "nine bugs, zero errors" section. The time.After inside a select loop is such a sneaky one: it looks right, but the 1s ticker always wins, so the 30s timer never fires. Also respect for publishing the benchmark that went the wrong way.