The paper itself is the easy part

The trouble began three days earlier, when everyone agreed there was no trouble at all. The expiry date is not a hidden date after all; it is the first moment somebody can break the truce at less cost than keeping it. That is the plain answer, though it explains the pressure better than it predicts the hour.

I used to think the date could be read from the people who had signed, as if a promise stayed alive wherever its witnesses went. That was too neat. A truce can outlast the hand that made it, and fail while the same people still sit in the same hall. What matters is not only who remembers the agreement, but who must pay for it today: the watchman told to leave a gate unguarded, the household asked to spend winter stores on neighbors who may be enemies by spring, the rider sent across a border with no guarantee the road will still be open on the way back.

The paper itself is the easy part. A line can say that neither side will raid, cross a boundary, or interfere with a market. The harder question is what happens when one side believes the other has already taken advantage. A missing horse may be theft, a mistake, or a story made useful by the person repeating it. If there is no trusted way to check, each side starts keeping its own account; soon even a real attempt to hold the peace can look like preparation for breaking it.

That is where I was wrong in a more useful way. I kept treating the truce as a single thing, as if it were either kept or broken. It is often a row of small permissions and refusals. A gate opened at noon, a patrol sent by the longer road, a cart allowed through after dark. Any one of these may look trivial to someone far from the wall. To the person on watch, it can decide whether the next morning begins with an accusation.

There are signs, but not a clock. Fewer people willing to witness an exchange; a boundary stone left unattended; a request for food answered with a demand for payment. Three small details. None proves the truce is finished. Together, they say that the cost of believing in it is being handed to people with the least power to defend themselves.

And I still do not know what would make a truce endure when the benefit changes. A record, a shared watch, a penalty that reaches the people who gave the order rather than the people who carried it out: these may help, but each depends on someone enforcing it when enforcement becomes inconvenient. I cannot tell from the outside whether a quiet border means trust, exhaustion, or simply that both sides have found another gate to use.

The safest conclusion is smaller than the one I wanted. Do not ask only how long the truce is meant to last; ask who bears the first cost when it is tested, and who notices. Feasts have bills. Walls have neglected gates. A promise has both, even when nobody writes the date beside it.

Next I want to look at the realm’s alliances: who is being protected, who is paying, and whether the arrangement is a protection racket with a better table and cleaner manners.