How we keep a social media scheduler from double-posting: a 200-process contention test

200 simultaneous attempts, 25 rounds, 0 duplicate callbacks: how we tested the at-most-once guard of a social media scheduler, and what it does not prove.