The writing, the grading and the sources on the live site are already good, and none of them change here. What changes is the loop around them. The map stops being a picture and becomes the thing you build. Every decision moves money, latency and how often you get woken up. The answers you got away with come back two levels later and ask for payment.
Multiple choice is one of them, not all of them. Each kind tests something the others cannot.
Three routes, graded best / workable / doesn't hold, with the reasoning for all three. This is what the site does today, kept as it is and given a confidence stake.
Choose a component and put it on the map yourself. Getting DNS right stops being a sentence you agreed with and becomes a thing you placed.
Three debugging steps, no order given. Order is the skill here: tracing before guessing is the whole lesson, and a list of options cannot test it.
A dial for the backup interval, with data at risk, storage cost and restore time updating as you drag. There is no correct number, only a defensible band.
During the incident, click the component that is actually failing. Diagnosis before action, tested by making you point at something.
Every action writes a timestamped line down the left of the screen, next to what the system said back. By the end of the night it is a postmortem you can scroll.
Running cost, p99, pages a week. Nothing lets you improve all three, so the debrief reads "shipped it, $34 a month, 160ms, one page a week" instead of a score out of nine.
A workable answer is not punished at the time. It leaves an open mark on your system, and the mark is what fails later. Hand out a bare IP in level one and Sam is locked out of the incident in level three.
Maya, Rio and Sam already exist in the writing. Here they keep score of you. Dismiss Rio twice and his bug reports get less useful, which is exactly what happens in real life.
The debrief renders the diagram you actually built, with your costs and your open marks on it. Two players finish the same campaign with different pictures, which is the only reason anyone would post one.
It is three levels, not two hundred and fifty-one. The leaderboard is filled in by hand, the replay modifiers are buttons that do not do anything yet, and none of it saves. It exists to settle one argument: whether the loop is worth building before anyone writes more content.