Debugging has always been detective work. Twenty years of production systems taught me that the resemblance is not decorative: there are suspects, alibis, red herrings, and a clock that charges by the minute. The only difference is that nobody hands you the genre's rules. This case file does. Read the brief, spend your minutes carefully, and remember the detective's real move: not "who looks guilty?", but "which question kills half the story?"
- A real class of incident, dressed as a mystery: the order existed, the ledger disagrees, four suspects are waiting.
- Every question costs downtime minutes. Brute force loses to the clock; a good question order wins comfortably.
- No dice anywhere. The case is deterministic, solvable, and the debrief at the end shows what each clue was really teaching.
14:02, mid flash-sale. A customer's order confirmed at checkout: payment approved, confirmation screen shown. Eight seconds later, customer support opens the ledger to help them, and the ledger says the order never existed. The customer has a screenshot. The ledger has nothing. Four components had the opportunity. Find the one that made the order vanish.
The board is yours, detective. A wrong accusation costs 4 minutes of embarrassment. Running out of budget means the pager wins. If you want a professional's opening move, think about which single question splits the pipeline in half.
How to Read a Room Full of Suspects
Without spoiling the culprit: the method that wins this case is the method that wins real incidents, and it has three moves. Eliminate: an alibi that clears a suspect entirely is worth more than a clue that makes another look shifty. Halve: the professional's opening question is rarely about a suspect at all; it is the one that splits the whole pipeline into a half that matters and a half that no longer does. The same instinct that finds a bug in a thousand commits in ten steps works on architecture too. Falsify: when a theory starts feeling right, the discipline is to ask the question that would prove it wrong, because evidence gathered after you have fallen for a theory has a way of agreeing with you.
And one more, free of charge, learned at more 2 a.m. bridges than I care to count: the loudest symptom is usually a witness, not the criminal. Something in this case is making an enormous amount of noise. Noise is not guilt.
Why the Suspects Are These Four
The cache, the replica, the queue, and the retry storm are not inventions for the game. They are four of the honest trade-offs every growing platform signs up for, the exact ones catalogued on the scaling ladder: every one of them buys real capacity, and every one of them opens a new way for the truth to briefly disagree with itself. A system that can vanish an order for eight seconds is not a badly built system. It is a normally built system, observed at exactly the wrong moment. Knowing which trade-off produces which illusion is most of what people mean by seniority.
The clock does not punish wrong answers. It punishes lazy questions.
Its Place in the Cabinet
The Case Files are the Cabinet's second drawer: mysteries that train the reasoning this craft actually runs on, the way the shelves train its ideas. This one exercises the Eliminator and the Skeptic, two of the six bands on the Ladder that is being built alongside these cases. If you solved it under fifteen minutes, you already interrogate systems the way the profession's best debuggers do. If you did not, replay it once with the halving question first, and feel the difference a single well-chosen question makes.