Dr. Park came to the patient room at 0820 on day eighty-six.
Not the clinical composure that assessment sessions produced. The physician expression Sora had cataloged three times in seventy-eight days: the one that preceded information that couldn't wait.
"The clinical safety notice was submitted at 1630 yesterday," Dr. Park said. She sat. "The sublevel operations oversight committee received it and issued an acknowledgment of receipt at 1800. Standard processing time for safety notices at this facility is forty-eight hours before a response is required."
Forty-eight hours. Day eighty-seven.
"The response this morning came at 0612."
Seven hours. Not forty-eight.
Sora waited.
"The oversight committee's response to the safety notice was routed through Director Kwon Mirae's division before reaching clinical administration." Dr. Park's hands on her knees. Flat. "The response classifies the instability in the sublevel subject's architecture as an anticipated research outcome rather than a clinical emergency. It authorizes continued monitoring under the existing protocol and indicates that no modification to the intervention schedule is warranted."
Anticipated outcome.
The research program had known the oscillation would reach this threshold. The biweekly applications generating the resistance-suppression cycle, the accumulation over eight months, the destabilization of the channel substrate at the developing node's junction sites — this wasn't an unforeseen complication. This was what they'd been measuring.
"They designed the intervention to produce this outcome," Sora said.
"I can't confirm the research program's intent from the oversight response." Dr. Park's voice careful. The physician's language staying within what the documentation could substantiate. "What I can tell you is that the response was processed through the Director's division and returned in seven hours, which indicates the situation was already known to and assessed by that level of administration before my notice arrived."
The research program's seventeen applications over eight months. Each one reinforcing the suppression. Each one triggering the counterclockwise resistance spike. The oscillation accumulating as an expected variable. And someone — Im Byeongsoo, or whoever had designed the protocol — had planned for the instability to reach this exact threshold.
Watching what happened when a partially mutated healer's suppressed architecture destabilized under sustained artificial pressure.
The retained subject as an experiment in induced architectural failure.
"What does destabilization produce," Sora said. Not a question about the retained subject specifically — a clinical question about what happened to healer-class channel architecture when the suppression-resistance oscillation exceeded the substrate's stress threshold.
Dr. Park was silent for four seconds.
"In the documented literature on class architecture failure — which is a small literature — the channel substrate under sustained mechanical stress develops micro-fractures at the highest-load sites. In the case of a developing node's junction pathways, the highest-load sites are the connections between the developing node and the established architecture." She paused. "The fractures develop gradually. At some point the developing node's channel connections lose structural integrity. The node's development terminates."
"The second node fails."
"If the oscillation continues without intervention. Yes."
The retained subject's architecture — the second node at below thirty percent density, the counterclockwise residual fighting against seven months of clockwise suppression, the oscillation already past stable threshold on day eighty-five. The research program watching it approach failure and classifying it as anticipated outcome.
"Dr. Park." Sora's voice. "What does an architecture look like after a developing node fails."
The physician's assessment: professional. Controlled. The affect management working.
"The established nodes are unaffected. The failed node's channel connection sites close — the substrate contracts around the failure, forming a scar at each junction. The healer retains full function of the established architecture. The developing node and all its potential development — lost."
Permanent.
The research program was conducting a controlled experiment in induced architectural failure on a retained subject with eleven months of continuous clinical data. They would observe the failure, document the channel substrate's response, and use the endpoint data to characterize what happened when a partially mutated healer's emerging node failed under artificial suppression conditions.
And Sora's safety notice had prompted a seven-hour administrative response that authorized the experiment to continue.
---
Minho came at 0800 day eighty-six and received the update from the slip she'd prepared. She watched his face while he read it.
He set the slip face-down on the table.
"The window," he said.
"The monitoring band is at clinical baseline sensitivity. Any counterclockwise output above threshold registers." She paused. "But the retained subject's instability is generating its own signal. The oscillation is producing irregular output in both the clockwise and counterclockwise ranges. If I time a pulse to the instability's counterclockwise spike, it would be masked by the subject's own output variability."
"Using their own architecture's resistance as cover."
"A short-duration input, below the spike amplitude, during the counterclockwise phase of the oscillation." She looked at the monitoring band. 0.08. "The band would see my output as continuous with the sublevel signal it was already detecting. A statistical artifact, not a discrete event."
"What does the input do."
"I don't know." The honest clinical answer. "The theory: the counterclockwise spike in the residual is the architecture's resistance response. An external input in the matching frequency range, arriving during the resistance phase, could reinforce the resistance rather than triggering the resonance cascade. The underlying healer-class resonance at 8% amplitude — the input might amplify it. Not to functional level. But enough to provide structural support at the junction sites where the fracturing begins."
"Enough to stop the failure."
"Enough to slow it. Maybe." She paused. "I would be putting counterclockwise energy into an architecture I've never directly engaged with. The monitoring band would flag a deviation if the timing is off or the amplitude exceeds the signal masking threshold."
"What happens if the band flags it."
"The enhanced monitoring protocols under the extension trigger an immediate response protocol. Review of the twelve-hour output record. Detection of a discrete healer-class emission event."
"And then."
"The courtyard exercise access is suspended. Movement restrictions escalate to room confinement. The committee reviews the flagged event as an unsupervised mana application." She paused. "The extension's administrative foundation becomes grounds for a reclassification recommendation."
Permanent monitoring facility. Kwon Mirae's division's custody.
Minho sat with this for several seconds. The S-rank hunter's risk calculus running clean.
"What does the retained subject have if you don't try," he said.
"Day ninety-seven. The next scheduled application. The oscillation's current trajectory puts the junction fractures at structural failure threshold inside eight to ten days. Day ninety-three to ninety-five." She paused. "Four to six days before the scheduled application."
The node fails before the research team does another intervention. And the research program classified this as anticipated outcome. They were waiting for day ninety-three.
"The exercise window tomorrow," Minho said. "Day eighty-seven."
"The instability should be producing maximum oscillation amplitude in the next twenty-four to forty-eight hours. If the resistance spike reaches a high enough amplitude, the signal masking window expands."
He absorbed this. Looked at his forearms — the compression sleeves, the suppressants managing what they'd been managing for three years, the partial repairs in the nerve tissue holding their incremental functional dividend through days without the forward healing session.
"What do you need," he said.
"I need Dr. Shin's most current estimate of the oscillation's counterclockwise spike frequency. The resistance phase timing. And I need you to be in the building when I do it."
He didn't ask why the last one.
"I'll be in Conference Room 1 at 0900," he said.
---
Day eighty-seven. Exercise window. 0912.
The northwest quadrant. The ventilation intake. The monitoring band at 0.08.
The passive reception running the sublevel two signal — and the oscillation was worse than day eighty-five. The counterclockwise resistance spike: not intermittent now but cycling at sixty-second intervals, the architecture's internal pendulum accelerating as the substrate stress increased. Each spike reaching higher amplitude before the adaptive clockwise baseline pulled it back down.
At 0924, the spike hit 22% of baseline amplitude.
She timed it. Forty-second measurement of the spike's duration at peak. Ten seconds of usable window.
She prepared the output. Not the forward healing rotation she'd used on Minho's tissue — not the broad cellular-interface application. Something narrower. The counterclockwise rotation directed at a precise frequency: 0.27 THz, the low end of the healer-class resonance range. Below the adaptive baseline's modulation floor. Targeted at the 6th node's directional axis toward the ventilation intake.
The amplitude: below the spike's 22% level. Masked by the architecture's own resistance output.
The next spike arriving.
She sent the pulse. Four seconds. The monitoring band registered 0.09 during the application — one point above baseline — then 0.08 as she withdrew.
The band didn't flag a discrete event. The sampling interval caught 0.09 and filed it in the noise range.
The sublevel two signal shifted.
Not the way she'd predicted. The counterclockwise spike that had been cycling at sixty-second intervals stopped cycling. The adaptive baseline's clockwise suppression frequency dropped sharply — from 0.31 to 0.26 THz, falling through the modulation floor she'd thought the architecture maintained as a lower bound.
The 6th node's passive reception, still running, tracking what her pulse had done.
The counterclockwise input had reached the underlying residual. Had amplified it. For approximately three seconds, the retained subject's architecture's counterclockwise resonance had registered above 30% of baseline — briefly, before the adaptive response compensated.
The compensation was the problem.
The adaptive baseline — the clockwise suppression that the architecture had been maintaining as its own resting state for seven months — had responded to the amplified counterclockwise input by accelerating the clockwise response. Not the fourteen-day biweekly cycle. The internal adaptive baseline cranking up the suppression to compensate for the brief counterclockwise amplification.
The architecture was suppressing itself harder.
And the junction sites under the elevated stress — the channel substrate fractures she'd been trying to prevent — responded to the sudden spike in suppression amplitude with a structural response that the passive reception registered as an abrupt change in the signal's character.
One fracture completing.
The node's connection to the established architecture: one pathway, severed.
She'd reinforced the resistance and triggered a compensatory suppression surge that exceeded the junction's load tolerance. The opposite of what she'd intended.
She went inside at 0941 and sat in the patient room while the monitoring band ticked at 0.08 and the ceiling tiles held at sixty-three and the healer on sublevel two carried one fewer channel connection than they'd had an hour ago.
The countermeasure she'd underestimated wasn't the research team's protocol. It was the architecture itself. Seven months of clockwise suppression had trained the biology to defend the suppression state. The architecture had fought the modification for eleven months, yes — but fighting something long enough changes you. The resistance had become load-bearing. Disrupting it had done more damage than leaving it alone.
She'd given the architecture proof that something external was trying to help it.
The architecture had interpreted that proof as a threat.
