Chapter 2: Eleven Seconds Local

Cold light crossed the recorder without touching its seals. It fell from the inspection bars above Recovery Bay Nine, passed over the dark casing, and broke against the raised edges of the evidence cradle. The serial marks remained visible in shadow.

Mara and Vale stood on opposite sides of the work surface. Neither reached across it.

Each inspected the write-disabled reader the other would use. Mara verified the first reader’s physical bridge, fixed contacts, certification strip, and hardware number. Vale repeated the checks on the second without consulting her notes. Then they exchanged devices, examined them again, and returned them to their original sides.

No instrument, operator, or inspection would be allowed to carry the result alone.

Mara connected through the first passive source and captured the recorder’s fixed witness window. The reader displayed no write path, repair prompt, or recovery control. Across the table, Vale acquired the same window through a second passive source set at a different optical angle.

Noise gathered differently in the two captures. A pale interference bar crossed Vale’s frame and disappeared when he corrected the mount. Fine grain shifted around the contacts in Mara’s image.

The recovered sixty-eight bytes remained identical.

Without moving the recorder, they exchanged source paths. Mara captured through Vale’s optical channel; Vale used hers. The noise changed again. The bytes did not.

Vale brought the four raw captures together. “Lexical services off.”

Mara checked his processing chain. “Confirmed.”

He had disabled automatic normalization, character repair, common-symbol libraries, controller placeholders, and recovery dictionaries. Two canonicalizers, installed from separate certified sources and sharing no repair tables, processed one capture from each path.

Both produced the same field boundaries and values.

The relay cache returned the frame unchanged. Its timing witness placed acceptance through P9-C eleven seconds before lockout. One input had been recorded, with no corresponding output increment.

Vale set that record beside the relay’s mechanical power witness. Unlike the controller log, it reported physical contact position rather than software state. Throughout the certified drift range, the outbound contact had remained open before, during, and after the frame entered the cache.

The request had reached the relay. Nothing had left through that path.

Probe Nine’s installed-controller witness preserved the matching transmission event.

Mara read the recovered text from her reader and then from Vale’s.

WITNESS/MV-17 PRESERVE RETURN ACCOUNT DO NOT COMPLETE RECOVERY WITHOUT WITNESS

Air moved beneath the evidence table with a low regulated hiss. Vale did not ask what the words implied. Instead, he opened the restricted registry in query-only mode.

The first search ran from designation to person.

MV-17 resolved only to Mara Venn.

Vale reversed the query, requesting every designation in the relevant restricted class associated with Mara. The registry returned only MV-17. Including retired records did not change either result.

The designation had been withdrawn nine years earlier. It was absent from active recovery directories, controller placeholders, and generic dictionaries. Mara confirmed those absences through her own reader, testing the exact code, spacing variants, and its stored registry form.

Nothing available to the recovery environment could have supplied MV-17 as routine completion.

Vale opened his corruption models. Starting with single-bit displacement, he widened the permitted damage through substitutions, shifts, contact echoes, partial overwrites, and repeated fragments. He tested the optical captures separately, then compared them with the cache and installed-controller witness.

None of the models produced the same sixty-eight bytes across all sources while preserving the counters, timing, and open-contact state.

He reordered the inputs and ran the challenge again.

The result held.

Mara opened a finding field. The recovered language invited a larger claim than the evidence permitted, so she entered the narrow one.

The request was transmitted through Probe Nine’s installed controller and deliberately selected Mara Venn through MV-17.

Vale tested each term against the record: transmission against the controller witness, routing against P9-C and the cache, selection against the exclusive bidirectional registry result and the designation’s absence from available completion sources.

“The selector is not established,” he said.

“No.”

“Nor that the selector was Nine.”

“Correct.”

Mara added the limitations beneath the finding. The evidence established neither authorship nor cognition, motive, truthfulness, personhood, continuing identity, or legal status. It did not show who or what had made the selection, why Mara had been selected, or whether the request’s account was accurate.

Vale reviewed the limits longer than the finding.

Then he countersigned Finding MV-17, fixing the conclusion and its boundaries into the same record.

He closed the signature field and opened the operational ledger.

“Eighteen labor-hours diverted,” he said, attaching the duplicated acquisition work, independent verification, and preservation staffing.

Both certified readers were marked unavailable while held with the evidence. Recovery Bay Nine’s occupation displaced the next intake slot, and the tug inspection moved by one shift.

Vale assigned every cost to Mara’s hold. None altered its duration.

The original six-hour preservation period had begun three hours and forty-three minutes earlier. The legal clock had therefore reached two hours and seventeen minutes remaining through elapsed time alone. Vale entered that calculation separately from the resource charges and turned the display toward her.