Chapter 28: The Gap After Restoration

The pump’s vibration climbed through Mara’s boots every four seconds.

It entered through the deck as a blunt pulse, faded along the rail, then returned. Condensation trembled on the exterior line. Beneath Orra’s inspection lamps, six seals crossed HL-B’s cabinet seam, their authentication threads intact.

“Plateau holds,” Orra said. “Three hours before the next material threshold.”

The thermal trace moved inside a shallow band. Exterior circulation was still running, but the measured temperature had stopped declining. No renewed cooling command had been issued. Nothing inside the cabinet had been addressed.

Vale watched another pump cycle pass through the rail. “Before we certify the trace comparison, the bridge language has to be clean.”

Mara looked at him.

“The draft says it was reduced to a preservation setting,” he said. “There is no such setting.”

“No,” Mara said. “There isn’t.”

Orra brought up the approximately 14:28 intervention record. The selector remained visible beneath its transparent custody hood, fixed in the service detent. No bridge crossed its actuator now.

Mara laid two fingers against the rail. Cold metal pressed back with each pulse.

“Reconstruct it from contact to withdrawal,” she said.

The intervention images advanced one authenticated frame at a time.

First, the Continuity halt had opened the scheduled relay and stopped branch circulation. Automatic recovery had failed because the local selector remained in inhibit. Orra’s preservation team had then seated the temporary bridge against the exterior actuator. Its load curve rose as it drove the selector out of inhibit.

The next frame showed the selector entering the service detent.

“The detent accepted full load here,” Orra said. “Independent mechanical retention confirmed by the actuator rebound.”

She advanced again. The bridge load fell to zero. Its contact arms separated from the actuator, and a later frame showed it completely withdrawn beyond the instrument boundary.

Installed circulation continued.

Pressure returned through the branch. The exterior line hissed, vibration entered the rail, and the authenticated temperature trace began its decline while the bridge was no longer present.

“The bridge actuated the selector,” Mara said. “It did not remain installed, retain the selector, or maintain circulation.”

“Agreed,” Vale said.

Orra entered the corrected finding:

The temporary preservation bridge moved the external selector out of inhibit. The selector entered and independently retained the service detent. The bridge was withdrawn. Exterior circulation continued through the installed system without the bridge.

She added a second line at Mara’s instruction:

Neither the bridge nor any continuing bridge setting maintained the subsequent thermal plateau. No continuing bridge setting exists.

Vale read both statements before touching the preliminary countersign field. “Accepted.”

Only then did Mara authorize the trace comparison.

The scope remained narrow: non-contact examination of exterior control surfaces and already exposed interfaces. No cabinet opening, interior imaging, diagnostic handshake, activation, simulation, software bypass, or destructive sampling. Access to every sealed surface required Mara and Vale jointly, with Orra as technical witness.

Orra read the limits back and positioned the magnification rig over the collar.

The authenticated 14:28 image still held the narrow reflective crescent first preserved during the intervention. Calibrated reprocessing of that unchanged image suggested two opposed compression contours and a release contour, but the underlying observation remained only a crescent or score.

Three bright grains sat embedded near its inner edge.

The exclusion hood over them stayed in place. Orra lowered the optical head until its range marks aligned outside the hood, then recorded distance, angle, illumination, and polarization.

At higher magnification, the collar’s coating ended in a shallow contact land around the selector access. One compression contour crossed that land almost squarely. The opposing contour ended near the recessed edge, where the three grains had lodged.

Vale leaned closer without touching the rail. “The spacing could take opposed jaws.”

“Could,” Mara said. “Not did.”

Orra overlaid the registered HL-B tool geometries. Their certified contact faces were chromium-nickel. None used cobalt, vanadium, and ceramic binder, the composition already returned by three agreeing non-contact scans of the embedded grains.

The registered tools also approached the collar through a narrower axial channel. Their contact faces could reach the outer contour but not the opposed inner contour at the observed spacing without striking the surrounding housing.

“Registered set excluded?” Vale asked.

“From producing the complete observed combination in their certified configuration,” Orra said. “Not from every possible contact, and not from later contamination.”

Mara approved the distinction.

Orra removed the collar overlay and shifted the rig toward the local-interlock service interface at the junction face. Its custody seal covered the fasteners but left the recessed service mouth optically accessible. The oscillator housing below it remained closed beneath its authenticated factory seal and secondary preservation band.

Cold light entered the recess at a low angle.

The interlock interface did not carry a crescent like the collar’s. One side of the service mouth showed a shallow diagonal flattening. The opposite wall remained coated and uncompressed. At maximum non-contact magnification, the flattening’s angle fell within the broad tolerance of one collar compression contour, but its usable width was smaller.

Orra ran surface spectroscopy from outside the seal boundary.

A weak cobalt response rose above background on the flattened side. Vanadium did not clear the instrument’s reporting threshold. No ceramic binder signature appeared.

Vale studied the spectra. “Transfer loss could account for that.”

“It could,” Orra said. “So could unrelated cobalt contamination, coating depth, a different contact material, or instrument geometry. The result does not reproduce the collar trace.”

“Does it exclude the same implement?”

“No.”

“Support it?”

“Only in a limited sense. One contact angle is compatible within broad tolerance, and one element is present above background. The spacing, opposed geometry, and full material signature do not correspond.”

Mara watched him process the answer. Vale preferred failure trees, but he did not prune them merely because one branch implicated his own institution.

“One undocumented implement might have contacted both interfaces through different surfaces,” he said. “Or two implements might have contacted them separately.”

“Those remain live alternatives,” Mara said. “Neither is preferred by this examination.”

Orra fixed a rigid comparison sheet to the rail. Its three rows were visible under the witness camera.