Escalation and SLA debt

An incident hurts you twice. Once through what it does to the machines, and once through the clock — and the second one is easy to miss because nothing on the rack turns red for it.

The grace window

Every incident gets a grace period after it appears. Resolve it inside that window and there is no penalty at all. Past it, the incident starts accruing SLA debt: uptime drains by 0.1 points per simulation step for as long as it stays open, up to a maximum of 45 points from any single incident.

Forty-five points is deliberately enough to lose a shift on its own. One ignored incident on a quiet early night can take you under the 50% threshold without a single machine going down.

The window shrinks

Early on you have a full minute to read, think and act. By the finale you have twenty-two seconds — which is roughly the time it takes to run one diagnostic and type one intervention, and not much more.

Escalated incidents do not shield you

This is the part worth internalising. An incident that has escalated stops occupying a concurrency slot. It does not block new incidents from arriving.

That closes what used to be an exploit: before this rule existed, a player who simply did nothing would fill every slot with ignored incidents and coast. A 1800-shift measurement found idle play winning 60–90% of nights two through ten. Now an idle player loses in every single configuration tested.

Reading it during play

There is no dedicated SLA meter. What you see is uptime falling while the rack looks healthy — that mismatch is the signal. If the machines are green and the number is still sliding, something is open that you have stopped looking at.