Table nine orders at 20:05. The kitchen printer jams somewhere around 20:06 and nobody notices, because a printer that has printed correctly for two years does not get checked.
At 20:55 the table asks. The waiter checks the pass, finds nothing, and the order is keyed again — this time as a rush, in front of a section that is already at capacity. Two other tables slow down. One of them leaves a review about it.
A screen fails differently from paper. It updates live rather than emitting once, so an order that has not been acknowledged is visibly still there rather than silently absent, and a station that has stopped consuming tickets is obvious within minutes rather than at the moment a guest complains.
Item-level state helps as much as the medium. The grill sees what it is cooking rather than a ticket for a table, the bar gets on with the drinks independently, and a line that is recalled is recorded as a cancellation rather than quietly disappearing from a spike.
Why this matters to you: the cost of a lost ticket is never one plate. It is the compounding effect on a service that was already at its limit, and it lands on a Saturday because that is when the system is under load — which is precisely when a paper-and-goodwill process has the least slack.