The official account said the city had been under a total curfew on the seventeenth day of Frost, when the eastern gate was sealed and no one crossed the river after dark.

The account was printed in a broadsheet with a crest at the top and a seal pressed into the ink. It described a quiet night, a threat from beyond the wall, and the courage of the watch captain who had ordered the gate closed. It said the order prevented a disaster. It said nothing about the people who had found the gate open before midnight.

A young verifier named Anja carried the broadsheet to the archive. The archive was a long room with shelves that leaned like tired trees. The head archivist, Soren, gave her a lamp and warned her not to trust a record merely because it sounded official. “Authenticity tells us where a record came from,” he said. “It does not promise that the writer saw everything, or told the whole truth.”

The broadsheet itself was real. Its paper, type, and seal matched other records from the same press. Anja traced its history through a custody register. The editor had received it from the watch office. The watch office had received it from the captain’s clerk. A second copy sat in the mayor’s papers, but both copies shared the same typeset plates. They were not independent confirmations. One copied account repeated in two places remained one source.

Soren opened a map of the river. The gate’s official closing time was 8:10 p.m. A mill roll, however, listed the night watch ending at 7:30. The miller had paid a worker named Bera for repairing a loom after a flood, and the payment was recorded in a large hand at the bottom of the page. The sheet had stayed in the mill’s custody until a flood-damaged chest was opened last winter.

It had a stain and a missing corner, but its original order was visible: the payment line came before the next morning’s total. Anja wrote the roll beside the broadsheet. A shorter workday did not by itself prove the gate was open. The mill could have closed early, and a worker could have crossed before the curfew. She needed context.

The wage book gave some. It recorded Bera’s hours, a lamp fee, and a meal at the worker’s table. The book had been kept by the mill owner, who had reason to make the work appear orderly. Still, the book’s pages and numbers appeared to have been written over time rather than copied for the official account. It was a different kind of record.

It agreed with the roll about the repair, but the two documents had been created for the same workplace and may have shared the same memory. Anja marked them as related evidence, not independent witnesses.

A private letter lay in a family chest. Bera’s cousin wrote that the gate was open when she passed it on her way home. She described lanterns on the bridge and a watchman who waved her through. The letter was written three days later, when fear had softened into detail. The writer had not witnessed the captain’s decision. She had seen one crossing. Her account was limited, but it was not made by the same office as the broadsheet.

The archive team compared ink, paper, and sequence. They did not ask only whether a document agreed. They asked who had handled it, when it entered the collection, and whether pages were added later. A loose sheet could be genuine and still be out of order. A ledger could be correct in its entries while missing transactions. An archivist’s later label was evidence about the archive, not automatically evidence about the event.

The watch office produced a second report. It said the gate closed early to protect the bridge from looters. A clerk named Vale had written a note at 8:40 p.m. The report listed three arrests, but the arrest ledger contained only two names. The missing name did not prove a third arrest. The page could have been removed, the number could have been copied incorrectly, or the report could have overstated the night. Anja recorded the gap instead of filling it with the most dramatic possibility.

The mayor’s secretary offered a copied list of curfew exemptions. The list seemed to prove that no one crossed because every traveler had a documented permission. Anja checked the source notes. The list had been copied from a lost draft, and the draft’s page was missing. A second copy had been made by the same clerk. The apparent agreement between the list and the broadsheet was circular: the documents had traveled together, not independently. Corroboration needed sources that could fail separately.

Anja assembled a table. The mill roll placed the mill’s night-watch end before the official closing. The wage book showed the repair was paid. The private letter placed a traveler near the open gate. The official account described a sealed gate and a threat. None of the records alone demolished the account. Together they made it necessary to ask whether the order, the recorded time, and the claimed effect were each being described by the same interested office.

Soren brought out a map of the original gatehouse. The lock mechanism was preserved, and the surviving tally marks showed a bolt drawn at different times. The marks could be maintenance marks rather than a record of the curfew. A tool fragment lay in the lock, but it did not show when it was placed. The physical evidence was valuable precisely because it did not supply a convenient story.

The team searched for the watch captain’s notebook. It was not in the archive. A rumor said it had been burned during the winter repairs. A rumor is not a record. The absence of the notebook left a silence that could hide many things, and the team marked that silence rather than claiming it proved destruction or innocence.

Anja asked the captain’s surviving apprentice to describe the night. The apprentice remembered a woman and child sheltering under the bridge, but also remembered shouting guards. His memory had been shaped by years of telling the story. The team wrote the recollection, the uncertainty, and the possibility of later influence. Memory could corroborate a detail while failing as a complete account.

The verification team prepared an interim report that refused the satisfying phrase “the broadsheet is false.” The sheet was authentic, and its account was incomplete and possibly wrong about the gate. The mill roll and wage book were related workplace sources. The private letter was independent but narrow. The copied exemption list repeated one draft. The missing notebook and arrest entry were unresolved gaps.

The team’s conclusion was narrower than the accusation and stronger than a rumor. The evidence did not support the claim that no one crossed after dark. It did support Bera’s work at the mill and, independently, a crossing seen by a private writer. The categorical no-crossing claim required qualification; the exact closure time and claimed safety effect remained unverified. The records did not identify every person at the gate, and they could not determine the captain’s entire intention.

The team next interviewed the families whose names appeared in the mill records. One woman remembered the flood and the loom, but not the gate. A boy remembered lanterns reflected in the water and thought they had been carried by a ferry. Their memories supplied color and possible details, not independent timestamps. Soren asked who had first heard each story and whether the family had repeated it after the broadsheet appeared. Some memories had been shaped by the official story; some had been shaped by fear. Provenance applied to speech as well as paper.

They compared the mill roll with the city’s bridge repair ledger. The ledger did not disprove the mill roll, although its copied timestamp could not independently verify the gate’s closing time. It came from a different office, but the clerk had copied the time from the watch captain’s daily return. The documents were therefore not independent confirmations. Anja drew arrows from one to the other. The map made the argument visible: two pages could disagree while sharing one ancestor, and two agreeing pages could still come from one error.

Anja’s team checked the original order of the mill book. The page numbers matched the neighboring pages, and the ink had not been removed where a later hand might have inserted a line. The evidence made the payment more credible, but the book still recorded what the miller wanted remembered. The team compared the amount with payroll totals, tool repairs, and a receipt for lamp oil. Corroboration grew through independent questions, not through repetition.

At the public reading, the captain’s advocate argued that the broadsheet must be true because it bore an official seal. Anja answered that the seal authenticated the office, not every statement in the text. She showed the custody path, the copied list, the private letter, the workplace records, and the missing notebook. The report did not turn uncertainty into a dramatic accusation. It named what the records supported, what they could not settle, and what new evidence would be needed.

The missing day remained missing, but the way the team described it was no longer missing precision.

The team then rebuilt the gate’s timeline from custody, not from the official account. The mill roll’s page stayed with the mill chest, but the wage book had been rebound years later. Anja compared the page numbers and binding holes before treating the entries as a continuous sequence. The books agreed about the repair, yet the wage book omitted one meal and the mill roll omitted a lamp. Those omissions did not prove either book false.

They showed that workplace records were made for work, not for a court watching a gate. Anja found a second private letter in a box of seed packets. It mentioned the bridge lanterns but had been written six weeks after the curfew, by a cousin who had heard the story from the first writer. The letter could corroborate that a lantern story circulated, not that the gate was open at a particular minute. The team marked the dependency and asked the family for the original envelope.

Its postmark placed the letter in the city, but a postmark recorded mailing, not the moment the writer saw the river. Uncertainty did not make the letter useless; it changed its role.

They tested the official account against a list of households receiving flour that night. The flour was distributed at the old bridge, and the ledger showed three families collecting after the stated closure. A clerk had reused the page for a later shipment, so the entries were not safely dated. The team photographed the page, compared its ink with adjacent entries, and kept it as a lead rather than a decisive fact. A dramatic account could be wrong in one place even when its paper remained authentic.

At the end, Anja wrote a short correction for the public notice: the curfew’s exact time was not established by the surviving evidence. The statement did not say that the gate stayed open all night, and it did not say that the captain had invented the threat. It marked the official claim as incomplete, the private crossing as plausible, and the missing notebook as unresolved. The team’s work had not restored the day.

It had given the missing day a boundary, a history, and a set of records that could be checked by someone who arrived after them.

Soren hung the broadsheet beside a document map rather than a verdict. The gate appeared in a web of custody, sequence, and dependency. Anja stood before it and understood that verification was not the hunt for one perfect paper. It was the patient separation of independent evidence from repeated copies, genuine records from incomplete ones, and silence from proof. The missing day had not returned whole. It had become a question that the surviving record could bear.