Chapter 58: Three Depths in Metal

The cart was absent. Its route was not.

ACK-2 and ACK-4 had already fixed the first and third post-retirement states of QR-CIV/6. Passive comparison had established a completed advance between them, and the retained sleeve’s corresponding witness space was blank. Mara allowed those premises onto the operations wall in four lines.

She refused to spend the recovery window proving them again.

“Pressure fields,” she said.

The technician replaced the chronology with three indexed maps of the key’s second tooth. Nothing behind the barrier moved. Fixed light had recorded the surface from twelve angles; the displayed model changed only by selecting among those authenticated images.

Orra leaned over the console, stimulant resin caught between her back teeth. Vale remained outside the forensic boundary with his hands folded behind him.

The middle advance had left two kinds of trace. The ratchet displacement recorded that the mechanism had completed its turn. The tooth recorded how force had entered and left it.

Mara enlarged the latter.

An aborted seating would have loaded the forward face and withdrawn along approximately the same line. A recording fault could duplicate a state value, but it could not raise metal, polish an edge, or leave directional drag. Routine verification without release would produce pressure at the stop and no unloading track beyond it.

The preserved trace did something else.

Force entered low on the tooth, climbed under rotation, settled against the stop, and then reversed across a narrower contact before withdrawal. The return path did not simply retrace the entry. Pressure had shifted after the mechanism released.

Orra stopped chewing. “It came back out after the latch cleared.”

“Yes.”

“Could a bad seating bounce through the stop?”

“Not with this unloading direction.”

Mara placed the service exemplar’s passive profile beside the evidence. She did not overlay them. The comparison needed differences as much as resemblance.

On the exemplar, a completed release-and-reseal cycle produced four regions: entry compression, rotational drag, a short unloaded interval after release, and withdrawal polish when the mechanism was reseated. QR-CIV/6 preserved all four. Its withdrawal region was narrower and slightly off-axis, but it occurred on the correct side of the released stop.

Vale moved closer to the wall. “How much of that depends on the state-ring sequence?”

“The timing depends on the authenticated sequence. The classification does not. The pressure profile is physical and independent.”

“So state the classification without attribution.”

Mara looked at him. “I intended to.”

She dictated the finding.

“The unidentified middle operation was a completed release-and-reseal cycle. It was not a duplicated record, an aborted insertion, or a verification motion against an unopened seal.”

The recorder returned her words for confirmation.

Orra pointed to the narrow withdrawal line. “That is not the same width as the registered service head.”

“No.”

“What made it?”

“A tool with a narrower release contact, or the registered head applied off-axis.”

“Which?”

“The surface cannot choose between them.”

The asymmetry was useful even without a name. Whatever had reopened the seal had left a repeatable mechanical profile: narrow release contact, low entry load, and an off-axis withdrawal toward the outer edge. If the cart, seal housing, or an associated tool survived, those measurements could support comparison.

Mara added that as a preservation note, not a conclusion.

Vale read the finding before signing. “This establishes reopening after the recorded handoff.”

“And before the officers’ return,” Orra said.

“It does not establish an operator, office, or authorization,” Vale replied.

“It also does not preserve the declared chain,” Mara said.

No one disputed that.

The officers’ account contained no release-and-reseal in the interval. The key contained one. Their return could no longer close the custody sequence merely because Transit’s system had labeled the movement complete.

Vale signed the non-attribution language first.

Mara watched his credential settle beneath hers. Continuity had gained no accusation, but it had lost the protection of an apparently closed chain.

“Retained sleeve,” she said.

The technician brought up the existing scan rather than beginning a new examination. Its first and third service spaces contained authenticated witness foil. The middle space did not. Mara left that absence in the margin and enlarged the alignment mark.

At ordinary scale, it was one dark line across the sleeve and seal index. Magnification separated its lower arm into two boundaries. The visible line crossed a second mark beneath it, offset by less than a degree.

Orra bent until the barrier light caught the silver at her temples. “Same instrument?”

“Not established.”

“Same sleeve?”

“Not established.”

“Made twice?”

“Two marking events are supported.”

Vale studied the divergence. “Could handling after removal spread the lower arm?”

Mara changed the image angle. The darker line continued beneath the visible mark and emerged on the other side with a coherent edge. Smearing would have widened with the surface grain. This line retained its own direction.

“No. The first mark was already present when the visible mark crossed it.”

That was the admissible limit. The sleeve had been aligned and marked, then disturbed enough to lose that exact relation, then marked again. The evidence did not show whether someone reset the same sleeve or exchanged it for another. It named no hand.

Orra’s resin clicked once. “They opened it, disturbed the witness seating, and closed it again.”

“The key proves the opening and resealing,” Mara said. “The sleeve proves disturbance after an earlier seating. Do not force them into a single actor before recovery.”

Orra looked ready to argue. Then she drew a breath and nodded.