Chapter 66: Four Names Before Dawn

At 23:08, Eli stopped Bell from calling the next position.

“Before twenty-three ten,” he said, “we close the interval from twenty-one ten.”

He spoke from behind the evidence table, hands clear of the strips, logger, and cabinet controls. His suspension had not changed. He held no key, operated no instrument, and possessed no authority to direct the watch. What he could do was identify the hole that would remain if everyone treated the next event as though the preceding two hours had not existed.

Bell looked to Voss.

“Reconcile it,” she said.

The cadence after the authenticated 21:10 event had generated twenty scheduled six-minute positions. Bell read them from the sealed logger while Ortiz followed the feeder record and Voss checked the custody initials:

21:16, 21:22, 21:28, 21:34, 21:40, 21:46, 21:52, 21:58.

22:04, 22:10, 22:16, 22:22, 22:28, 22:34, 22:40, 22:46, 22:52, 22:58.

23:04.

Those were the nineteen intermediate positions.

The twentieth landed at 23:10.

Every position remained present in the full chronology. The logger showed no restart, clock loss, missing sample, range change, or configuration change between 21:10 and 23:10. Custody initials covered the same period without a gap.

Bell drew a bracket around the nineteen intermediate times.

“Preserved in the full cadence record,” he said. “Excluded from the limb-attribution run.”

The synchronized passive comparison package—the separate CR-4A and CR-4B traces, external feeder trace, return monitor, and vibration puck under the validated arrangement—did not begin as a complete set until 23:10. The earlier positions were therefore neither erased nor represented as proof that activity had ceased. They remained outside the narrower finding because they lacked an equivalent source package.

Eli wrote the distinction in full. Six later powered knocks would not become the whole night merely because they were the portion capable of identifying a limb.

At 23:10, Bell called the twentieth position.

“External rise.”

The external trace lifted cleanly. One point six seconds later, CR-4B moved.

CR-4A stayed flat.

The puck trembled near the custody stripe, and a single iron knock crossed the flooded floor. Its source sounded sharp; the chamber returned it hollow. Then the puck steadied, CR-4B settled toward its prior load, and the external trace fell away.

No one approached the cover.

The same order returned at 23:16: external trace, 1.6-second delay, CR-4B load, vibration, iron knock. CR-4A remained level.

Bell called each source separately. Eli recorded the calls with space between observation and conclusion.

At 23:22, the progression repeated.

The 23:28 position produced another external rise and another CR-4B response. The puck moved. Iron sounded through the water. Neither the cover nor the seam visibly shifted.

At 23:34, the fifth positive event arrived on phase.

After each event, Bell checked logger continuity. Nothing indicated that the recurring order had been manufactured by an outage or instrument change. He kept the external trace, both limb traces, puck record, return monitor, and logger separate. Their agreement was temporal, not the product of an interpreted overlay.

Voss watched Bell’s hands. Ortiz followed the accumulating times. Havel remained inside the marked witness position, representing the Flood District’s hazard interest without touching coalition equipment.

At 23:40, the sixth scheduled position arrived.

Nothing answered.

The external trace remained level. CR-4A and CR-4B showed no movement. The puck registered no vibration, and no knock crossed the floor.

Bell waited beyond the point where the 1.6-second response should have appeared.

“No feeder movement,” he said. “No vibration. No knock.”

Eli entered an absence at the scheduled position. He did not shorten the interval or move the sequence to hide it.

The logger had continued through 23:40 without restart, clock loss, missing samples, or a range change. This was a missing powered event, not a missing record.

At 23:41, the isolated return monitor made one mark.

Bell compared its position with the earlier record without touching the scale. “Same post-gap position as twenty-one oh-five.”

Eli attached both times to the comparison. The return indication followed the absent event, but timing alone did not make it the cause. It remained isolated from the cabinet file and outside the feeder sequence.

At 23:46, the external trace rose again.

“External,” Bell called.

CR-4B moved 1.6 seconds later.

“B load.”