Chapter 48: Thirty-Seven Seconds
The missing operation record lasted thirty-seven seconds.
The custodian found it while aligning Sterling's current safety log with the independent offline image. We watched from behind the boundary in the dual-source comparison bay. Text remained masked. Only version events, account categories, and synchronized time ranges appeared.
The two histories matched until the risk-summary record entered the sequence.
The custodian aligned them by the archive holder's receipt time. Before the disputed window, the event counts and protected version markers were identical. After it, both histories displayed the same later version. The difference existed only in the operation connecting the two.
The offline image showed an operation beginning with one protected version and ending with another. The current log moved between the same versions without the intervening event.
Elapsed time: thirty-seven seconds.
The comparison did not reveal what either summary said. It showed that a version had been replaced during that window and that the offline image retained an operation record the current log did not.
Celeste called thirty-seven seconds too brief to matter.
"Duration does not decide significance," the custodian said. "The event type does."
The event type recorded a replacement, not an ordinary viewing or automated retention check. It did not show whether the replacement was proper, who requested it, or what text changed.
Nathan requested the exact clock time.
The chair denied public or unrestricted display but allowed the custodian to classify the event. It was not the 2:17 approval-field revision David Ren had identified. That event involved an approval field and a Safety Administration Service entry.
This was a different record, in a different sequence.
"Then it has no connection to the disputed approval," Nathan said.
"No connection has been found," Rhea replied. "The question is why the current safety log lacks an operation preserved in its independent image."
The custodian opened the role category, not the user identity.
LEGAL PRESERVATION ADMINISTRATOR.
It was not a research account. It was not assigned through the scientific team. The role table placed it inside the legal preservation function, where more than one authorized credential could have existed.
The role matrix showed that scientific staff could not grant themselves that category. The person using it would have needed an authorization chain outside my former team, but the comparison report did not contain that chain.
Celeste focused on that uncertainty. "So you cannot identify a person."
"The report names a role, not a person," I said.
She had expected me to turn the category into Nathan. I would not repeat Sterling's method with a different convenient name.
"You cannot identify intent either," she said.
"It does not identify intent."
"Then this is an archive difference, not evidence of wrongdoing."
"It is evidence of an operation record present in the independent image and absent from the current log."
The panel adopted that sentence.
It also directed the custodian to preserve both the full masked comparison and a narrower report for the license record. The first would remain inaccessible without participant authorization. The second contained only the thirty-seven-second interval, event type, and account category.
Mina attended through her lawyer. The chair asked whether her safety role authorized the legal preservation category.
"No," she said. "I could request preservation. I could not assign or operate that role."
Her answer eliminated her own authority. It did not identify whose authority remained.
The comparison team locked both histories again. No participant text entered the finding, and no one received instructions for accessing or changing the systems. The report described only the preserved difference.
Nathan argued that an ordinary migration could explain why one history retained an intermediate event and another did not. That possibility required a record. Five years earlier, the board's legal hold had already ordered preservation of system migrations and duplicate-deletion logs.
Rhea asked the custodian to execute that part of the existing hold.
The search did not revisit Meeting Forty-Four, the door logs, or the 2:17 service-account history. It targeted governance records for the safety-log archive during the period after the incident.
Near the end of the session, the custodian reported one responsive migration packet.
He preserved it unopened for source classification.
The packet had been issued after the incident.