Meridian’s response arrived several days later. It confirmed that the recovered message had been sent through a supervisory account associated with the restricted records department, but the company could not identify the individual who had composed it. The account was authorized to communicate task status to external recipients. Its credentials could be used from the shared workstation, and the surviving logs did not preserve enough detail to distinguish every person who had accessed the machine.
The response also clarified that the department’s internal closure code did not certify that a source comparison had been completed. It meant that the assigned task had reached an administrative endpoint. A separate result or transfer reference was ordinarily expected when the task involved verification of an original document. In the Northbridge file, Meridian could not locate either.
I read that paragraph twice. It was the clearest description we had received of the gap between what the department had recorded and what the insurer had understood. The task could be closed without a completed comparison, yet the message sent afterward had been treated as sufficient for processing. Meridian was now acknowledging that its own status code did not mean what the insurer might have believed it meant.
Evelyn pointed out that the response did not say the message had been sent in error. It did not say the sender had misunderstood the procedure, either. It described the system and the missing records, then stopped short of assigning responsibility. Meridian was admitting the limits of its evidence, not explaining why the assurance had been issued.
Detective Raines asked the company to preserve every remaining record associated with the supervisory account, the shared workstation, the Northbridge file, and the insurer’s mailbox address. He requested any retained access logs, message-routing data, backup indexes, and records showing who had been issued the relevant credentials. He also asked Meridian to identify whether the closure code had been selected manually or applied automatically when the task was marked complete.
The company replied that the code could be entered by an authorized user. The system might then carry the status into later reports, but it would not independently verify that the required source comparison had occurred. In other words, the software recorded what a person entered; it did not guarantee that the real-world action behind the entry had taken place.
That explanation made the missing result more troubling, but it still did not tell us who had entered the code. The roster showed that Adrian Cole had supervised the restricted archive during the relevant period. It also listed employees who had used the shared workstation. The authentication log showed activity under the supervisory account shortly after a clerk’s login, but the available data could not establish who had been at the keyboard when the message was prepared.
Cole had already denied knowingly suppressing a review. He had confirmed his supervisory role but could not remember personally examining the original, writing the note, or closing the task. Those answers left several possibilities open. He might have made a decision that he no longer remembered. Another authorized employee might have used the account. Or the message might have reflected a routine administrative action whose significance was not recognized at the time.
Raines warned us against treating uncertainty as evidence of guilt. A shared account was a weakness in the system, not a fingerprint. The roster could identify who had an opportunity to act, but opportunity alone did not establish authorship or intent. If the investigation was to identify the person responsible, it needed a record that connected a human being to the specific decision.
We returned to Simon Bell’s interview notes and compared his recollection with the new response from Meridian. Bell had remembered a task suspended pending source confirmation and an instruction not to release the original through ordinary intake. He had also remembered someone asking whether the original had reached the person responsible for reviewing it. At the time, he could not identify the speaker with confidence. He had never claimed to see the original or the final comparison result.
Raines asked Bell to clarify the setting of that question. Bell said he remembered it as a practical concern raised while the file was still being handled, not as an accusation that anyone had deliberately hidden a document. He could not remember the exact date or identify the person who asked. He was certain only that the question had been important enough to stay with him.
“Did anyone answer it?” Raines asked.
Bell thought for a while. He remembered discussion about the file’s status, but not an answer that named a reviewer or confirmed a completed comparison. He could not say whether the answer had been given in a conversation he had forgotten or whether the question had simply remained unresolved. He would not claim that silence meant the original was missing; he could only say he did not remember receiving a reliable confirmation that it had arrived.
His recollection did not identify the sender, but it fit the documentary sequence. Before the closure message was sent, someone had apparently questioned whether the original had reached the intended reviewer. After the task was closed, the insurer received an assurance that allowed the authorization to proceed. Between those events, the file acquired a status that could not be supported by a surviving comparison result or destination receipt.
Evelyn suggested that we focus on the person who had authority to resolve the question, rather than the person who happened to use the workstation. The restricted-archive note required the original to be held pending source confirmation. Someone had the authority to release it, document the result, or keep the task open. The critical decision might have been made before the external message was sent.
That meant we needed the full chain of internal instructions, including any records not formally attached to the Northbridge file. Raines requested copies of the department’s daily task sheets, handover notes, and supervisor communications from the relevant period. Meridian said some of those records had been retained in paper archives and would require additional review.
While the search continued, we asked Cole for a second interview focused on the newly clarified meaning of the closure code. He agreed through his representative. We did not accuse him of sending the message. We asked whether he had understood that an administrative closure did not certify source confirmation, whether he had authority to close a task while the comparison remained outstanding, and whether he could remember any instruction to communicate the closure to the insurer.
Cole said he understood the distinction between an internal status and a verified result. He also said that tasks could sometimes be closed administratively when an assigned process had been discontinued, transferred, or superseded. Such closure did not necessarily mean every substantive question had been answered. But he could not recall the Northbridge file being discontinued or transferred, and he could not explain why the insurer had been told that the authorization could proceed.
He repeated that he had not knowingly concealed a missing original. He could not remember sending the message or directing anyone else to send it. He acknowledged that the supervisory account was available to authorized staff, but he denied that this fact proved he had used it for the Northbridge communication.
Raines asked whether Cole remembered the question Bell had described: whether the original had actually reached the reviewer. Cole said he did not recall hearing that question. He did not dispute that someone might have raised it. He simply could not place it within his own memory of the file.
When the interview ended, I felt the same mixture of disappointment and unease I had felt after the first one. Cole’s answers did not resolve the case, but neither did they justify a conclusion that he had deliberately falsified the record. His role gave him authority over the process; the records did not establish that he personally made the disputed decision. It was possible that he knew more than he remembered, but possibility was not proof.
The paper archives produced a handover sheet from the restricted records department. It contained brief entries about pending tasks and several references to files that had been moved between staff. The Northbridge entry was incomplete, but one line noted that the source question remained unresolved at the time of handover. There was no accompanying instruction to cancel the comparison, no named reviewer, and no receipt confirming that the original had reached its destination.
The handover sheet was important because it placed the unresolved source question in the department’s records rather than only in Bell’s recollection. Someone had recorded that the question remained open. Yet the task was later closed administratively, and the insurer received the message that allowed processing to continue. The document did not identify who made the later decision, but it made it harder to argue that the uncertainty had never existed.
Meridian said the handover entry might have been superseded by a later action that was not preserved in the surviving file. That remained possible. The company could not produce the superseding instruction or identify a reviewer who had completed the comparison. The absence of those records did not prove that no one ever reviewed the original, but it meant that Meridian could not demonstrate that the required review had occurred.
I asked Raines whether the handover sheet changed the investigation’s direction. He said it strengthened the chronology. We now had a written record that the source question remained unresolved at handover, followed by an administrative closure and an external assurance. The next task was to determine who had authority over the file at each stage and whether any surviving communication explained the change.
Evelyn cautioned me not to let the word “unresolved” become a substitute for a finding. It was a description of the file’s recorded state at a particular point in time. A later review might have resolved the issue, but no surviving result showed that it had. Our strongest claim remained narrow: the records could not establish that the original had been compared with the authorization before the insurer proceeded.
That narrow claim was already consequential. The insurer had treated the authorization as acceptable, the supplemental report had relied on information collected under it, and the missing confirmation had not been disclosed as a limitation in the report. We did not need to prove who had used the supervisory account to show that the process had failed to preserve evidence of the required safeguard.
As we prepared the next request, I realized that the investigation had reached a point where the absence itself could no longer be treated as an inconvenience to be repaired later. A missing signature might be explained by a filing error. A missing attachment might be recovered from a backup. But here, every new record repeated the same pattern: a source question was open, a status changed, and the evidence that should have connected the two events could not be found.
We still did not know whether that pattern resulted from deliberate action, careless administration, or multiple failures that reinforced one another. What we knew was that the system had permitted a statement of closure to travel farther than the evidence behind it. The person who sent the message remained unknown, but the authority of the message had never depended on a named sender appearing in the surviving record. It had depended on the recipient believing that Meridian had completed the necessary work.
Click here to continue reading: PART 32: The Original Could Not Be Replaced by a Record of Its Own Disappearance
I Bought My Son a Laptop With Money Nobody Knew I Had Received, and By Morning My Family Had Begun Asking Questions I Couldn’t Answer
Part 31 of 35
