A counter has gone quiet with paid orders still arriving. Here is the order to check things in, and why nothing is lost while you do it.
A counter that stops printing is not a counter that stops selling. Orders are still being paid for and still arriving, and that is the moment people panic: it feels like every one of them is vanishing into the gap. They are not. A paid order is written down the instant it is taken, before anything is asked of a printer, and it stays written down until a ticket for it has come out of one.
So the first thing to do is nothing. Do not refund, do not ask guests to order again, and do not start writing names on a pad in case. The queue behind the counter is holding, and when the printer comes back the tickets it owes will print in the order they were taken. Your job is to find out which of three dull things is wrong.
That sequence is not arbitrary. Paper and a knocked-open cover account for most of it, power for most of the rest, and the network for the small remainder that people always assume is the cause. Checking in that order means you are usually printing again before anyone has had to think about networks at all.
Put the counter's items on another counter if there is one that can make them. One bar covering two menus for a few minutes is uncomfortable, but it keeps sales moving and gives you room to work on the printer without an audience. Tell the staff at the covering counter what has moved, and tell them when it moves back.
If nothing else can make those items, take them off the menu rather than letting people pay for something no counter is going to see. Guests will accept being told an item is off. They will not accept paying for one and then being sent somewhere else to argue about it. For the people already standing at the quiet counter, a name, an item and a sentence about where to stand is enough; silence at a counter reads as nobody is coming.
Watch the first few tickets come out and match them against what the console says that station owes. Anything the printer refused while it was down is held and can be sent again from the console, either one ticket at a time or as the whole failed set, and nothing you resend will print twice. If a guest has been standing there a while, hand their order over directly rather than making them sit through the backlog.
Then find out what the console said while it was happening. A station that drops off is flagged there and an alert goes out, so the useful question afterwards is not whether you were told but how long it took for somebody to look. That gap is the thing worth shortening before the next event.
Write down what it was and what you did, at the counter, on the night. Three events of that and a pattern appears: the same socket, the same shutter, the same roll nobody changes at the start of service. Most counters that go quiet twice go quiet for the same reason both times, and the repair is dull — a spare roll inside every counter, the printer on a socket nothing else gets unplugged from, and one person per shift whose job is to glance at the console rather than wait to be told.