The Level Four pulse arrived during testimony.
By then Elias had told the Sillglass story in five incompatible rooms.
Medical asked when Analysis narrowed his vision, whether he saw the transparent ridge before falling, and how long nausea lasted after the topology event. The answers turned his model into an exposure record.
Alpenkern Safety asked which abort criterion existed before the Glasswing probe and which one Klara created after it. They cared less about west being correct than about the moment east became unacceptable.
Navigation Engineering asked whether the SG-15 tag had ever reported a false coordinate. Elias said no three times. The coordinate remained correct. The connection failed. Each repetition sounded more like he was defending the software from his own conclusion.
The district observer asked who first used the phrase displaced corridor. Georg remembered Elias saying map problem. Tobi remembered one segment shifted. Helmet audio contained both at different times. The record had to distinguish a model developing under pressure from a perfect diagnosis announced at the first clue.
The works-council representative asked whether anyone continued work after the contradictions should have stopped the route. Klara answered for command. Georg answered for harvest. Elias answered only for the time he remained inside Analysis after useful comparison ended.
Each interview held one fragment. No room contained the story he remembered, because institutional truth was assembled from records that had different jobs.
The Sillglass escape and its immediate hazard objectives had already brought Elias to 674 / 900 integrated experience. Formal validation of the navigation model remained unresolved while the district review was open.
The System had not yet decided whether his observations constituted a resolved contribution or a plausible opinion attached to a crew that found its way out. Elias disliked that the blue total and the district finding appeared to be waiting on the same word: validation.
He feared the review would choose a simpler account. Tobi's probe found the passage. Klara selected it. The topology event proved the map unreliable. Elias could disappear into the sequence without any statement becoming false.
Elias had just said, for the third time, that he could not prove the one-corridor model before Tobi’s probe entered the side passage.
The district observer asked, “Yet you recommended the route.”
“I recommended testing the route.”
“Distinction noted.”
Elias had learned not to fill that pause. He placed both hands flat on the table and waited while the observer attached his answer to the timeline.
Blue text opened over the recording light.
Dungeon objective resolved: Navigation hazard identification and extraction support Integrated experience gained: 226 LEVEL INCREASE Level 3 → Level 4
Heat struck through Elias’s chest.
The pulse tightened every muscle already fatigued by six hours of testimony and interrupted sleep. His hearing sharpened around the recorder's cooling fan. The blue text remained legible while the faces behind it doubled.
He gripped the table.
The observer leaned back. “Is something wrong?”
“Level increase.”
The room changed procedure immediately.
Recording paused. Medical entered. Petra postponed the remaining questions. Elias spent the next hour beneath monitoring pads while the most important professional meeting of his life continued without him.
The district observer sealed the unfinished transcript rather than accepting answers during stabilisation. A clerk recorded the exact interruption time. Petra collected Elias's written exhibits but did not brief him on testimony he missed. Procedure protected the inquiry from his altered condition and him from being treated as available because the level looked like good news.
On the medical cot, the new level felt badly timed and undeservedly theatrical. Validation of his model produced 226 integrated experience. Farid's shield had held the plate. Tobi's fibre found the first viable passage. No blue window appeared over their names where Elias could see it.
F-Potential stabilisation took nearly twelve days.
The first two belonged to Medical. The next four belonged to restricted work and headaches that arrived whenever he tried to compare moving data. By day seven, his legs had adapted enough that stairs felt subtly mistimed. Strength had increased before coordination trusted it. Nika made him repeat gait tests he had passed at Level Three.
At home, the level was less dramatic. Elias could carry the laundry basket with less strain and still needed to remember not to take both flights at once. Lena asked whether Level Four made him four times as strong. Sophie asked what the denominator was. Anna asked whether he had taken the afternoon medication.
His badge updated automatically. His qualifications did not. Level Four did not make Sillglass extraction familiar, remove the imposed Analysis-release rule, or turn testimony into command authority. The largest immediate change was a mandatory medical calendar and three notifications from training providers advertising what a Level Four body might buy.
Elias deleted them. Stabilisation was not preparation just because both occupied time.
During that time, Sillglass data established the route failure beyond argument. The coordinate nodes had remained accurate while a topology event changed their connection. Elias’s physical contradictions had detected what the map architecture could not represent.
Navigation engineers reproduced the failure in a simulation using the recorded node states. The old system checked coordinate integrity, tag sequence, and route checksum. All passed after the edge changed because the connection was stored as a relationship between valid nodes rather than a separately expiring fact.
The Glasswing chamber existed behind the correct SG-09 coordinate. After the topology drift, SG-15 connected to it instead of the Gate corridor. The map had not lied about either place. It had told an obsolete truth about how they touched.
Tobi's cut probe fibre showed a second change during withdrawal. Its tension path crossed a seam that closed after the probe passed, confirming why following the first successful line at Junction Three would have led toward the colony. Elias's updated descent model matched the final physical map within one turn.
The final incident summary credited the crew.
Klara for halting and ordering verification. Tobi for probe confirmation. Farid for holding the topology gap. Nika for ward and medical coverage. Georg for historical harvest interpretation. Elias for assembling and communicating the displacement model.
His name occupied one line among six.
It felt better than occupying the page alone.
The line was enough because removing it would make the account less true.
Klara's credit was bounded. She wrote that Elias assembled the first viable displacement hypothesis and communicated uncertainty well enough to test. She also recorded that his extended Analysis caused a preventable stumble and contact noise, and that he required an imposed release schedule afterward.
Contribution did not require the report to make him flawless. It required his work to remain causal after the errors stayed visible.
Analysis advanced after the formal reconstruction.
ANALYSIS has reached Common Level 6. Communicated-model validation recognised. Confidence calibration improved.
Alpenkern paid hazard recognition and a navigation-correction bonus. The district adopted a software change requiring topology alerts to invalidate corridor connections rather than merely coordinates.
The bonus equalled a little more than one interior supplement, divided by the crew agreement. Farid received medical coverage for the compressed hand and replacement shield edge. Tobi's destroyed probe went to incident cost instead of his equipment variance. Georg's delayed pharmaceutical cases were graded without penalising the crew for chain time.
More important professionally, Sillglass added a navigation-anomaly qualification note to Elias's record. It did not grant command or site access. It made Operations request his review on two old route discrepancies before closing them as local error.
Then Petra offered Elias a desk.
It was a good desk.
Petra had him work at it for two stabilisation days before asking for an answer. The first task compared resin-downgrade reports from Sillglass with lamp settings and time between extraction and dark-sleeve closure. Elias found that one crew's apparent handling problem clustered around a batch of cases whose internal shutters closed slowly below ten degrees.
He did not enter a Gate. He prevented another crew from being blamed for equipment drift.
The second task involved Hachinger silverleaf assays. Small clusters from South Two had lost value across three weeks despite stable technique. Elias aligned collection footage, Processing calibration, and reference samples. The drop began on the morning Assay Station Four received a replacement sensor.
Technicians recalibrated it. Twenty-seven samples moved back into the correct grade.
The work produced decisions with a reach no single harvest shift could match.
The resource-analytics office had windows, normal air, and chairs designed for humans rather than decontaminated armour. The salary exceeded his harvester base by eight percent. Hours were regular. Interior rotation would fall to two collection-verification shifts a month.
On the second stabilisation day, Petra placed him beside Miriam Koch, a senior analyst who had spent eleven years in pharmaceutical collection before a shoulder injury ended her loaded-field clearance. Miriam's screens displayed three concessions and a queue of anomalies sorted by expected financial impact.
“Do you miss the crews?” Elias asked.
“Every week.”
“Do you regret transferring?”
“Most weeks, no.”
It was not the clean warning he had wanted.
Miriam showed him how analytics findings failed. A report could be correct and arrive after Operations committed resources. A model could overstate confidence and force a crew to disprove it inside. A desk worker who treated field notes as dirty data could erase the exact uncertainty that kept people alive.
“The office isn't outside the Gate,” she said. “It is farther from consequences. That makes discipline more important, not less.”
At lunch, she left on time for a physiotherapy appointment without negotiating her value first. Her calendar contained protected medical blocks, paid coursework, and one monthly field verification. The position was not exile. It was a profession Elias could respect.
The contract converted most shift and exposure variance into base pay. Anna's adjunct shortfall could close without early-route supplements if the fund remained delayed. Elias would reach home before dinner on four nights out of five. Medical expected fewer mana-pressure reviews. His knee and back would accumulate less load.
There were costs the spreadsheet could not express symmetrically. Lower exposure meant slower System progression. Two field verifications a month would maintain a harvesting licence but not build broad site competence. If Elias later applied for operational endorsements, he would be starting from a role designed to keep him behind the decision boundary.
Petra did not hide that. She also did not accept it as a universal loss.
“Progression is not in the job description,” she said. “Accurate work is.”
“For people with high Potential, progression becomes part of staffing.”
“For some roles. Not this one.”
The desk offered equality of pay through usefulness and inequality of opportunity through safety. Neither fact cancelled the other.
The analytics track also paid for statistics coursework and cross-site data access. Sillglass, Hachinger, and the Stonefold survey concessions would all feed his desk. Analysis had finally found a job whose bottleneck was evidence rather than attributes.
“You would compare field reports across sites,” Petra said. “Harvest anomalies, contamination, topology, assay mismatch. Analysis has demonstrated value beyond manual collection.”
“Who acts on the findings?”
“Operations, ecology, site teams.”
“Who collects them?”
“Field staff.”
“So someone else goes inside.”
“Usually.”
“Could I retain one regular crew rotation?”
“Not at this staffing level. Two verification shifts per month, scheduled where data needs field confirmation. They may be with Klara. They may not.”
“Harvester licence?”
“Maintained through minimum practice and medical eligibility. Your permanent position would transfer to analytics.”
The offer did not leave the Gate entirely. It changed fieldwork from the centre of his profession into support for the desk.
The office wall displayed Sillglass’s corrected route model. Elias could have prevented the same error across every Alpenkern site without carrying another frame.
Anna examined the offer that evening.
They compared three years, not one month. With analytics, the household could restore help, preserve insurance, and contribute modestly to savings without assuming a treatment-fund victory. With the mixed field rota Elias currently held, one injury or two cancelled supplements erased the annual margin. The expected-value argument was not close.
She entered the base pay into the spreadsheet, removed variable interior supplements, reduced transit meals, and restored two home-help visits they had postponed. The result left a modest positive margin before any treatment-fund award.
“This is the first offer that makes the budget boring,” she said.
“Boring is good.”
“Usually.”
She turned to the recovery schedule. “It also means you are here on clinic days without trading shifts.”
Sophie, doing homework at the other end of the table, looked up but did not join unless invited. The household had become better at keeping a decision with the adults who owned it.
“More money,” she said.
“Yes.”
“Regular hours.”
“Yes.”
“Less chance of being bitten beneath industrial equipment.”
“Statistically.”
“And you hate it.”
“I don’t hate it.”
“You look as if they promoted you to a prison.”
“It’s valuable work.”
“That is not what I said.”
Elias looked at the offer. Nine months earlier he would have accepted before Petra finished speaking. It solved money, reduced risk, and turned his mediocre starter skill into a professional identity.
It also placed glass between him and every Gate.
Elias tried to describe that cost without making danger sound noble.
“When I found the assay drift today, the work mattered,” he said. “I liked it.”
“That seems inconvenient for your prison argument.”
“It isn't a prison. That's the problem.”
Anna closed the spreadsheet. “Then do not refuse it by inventing reasons it is bad. If you say no, say no to the good version.”
The good version meant more money, more evenings, lower physical risk, professional education, and work he could do well. Refusing it would not prove courage. It would spend all of those advantages on something he had not yet named honestly.
“I need time,” he said.
Anna nodded. “Then take time. Stop trying to make the correct answer arrive before you know the question.”
At the terminal, Elias visited the analytics office twice during stabilisation. He reviewed Hachinger yield data and caught a calibration drift that had misclassified small silverleaf clusters. The work satisfied him.
Then Klara’s crew returned from a routine route without him.
Their armour carried blue mud. Tobi complained about the western repeater. Georg brought three samples nobody had classified. Farid had a new scratch across his shield.
Elias stood on the clean side holding a tablet.
The separation hurt in a way danger did not explain.
Petra found him there.
“The offer is not punishment,” she said.
“I know.”
“It is the best use I currently have for your skill.”
“I know.”
“You may refuse.”
That was why the choice had weight.
“If I accept later?” Elias asked.
“The position may exist. This offer may not. Miriam's queue is growing now.”
“So time matters.”
“Time always matters. That does not make every deadline coercion.”
Petra gave him forty-eight hours and returned to Operations. Elias remained at the clean-side window while Klara's crew crossed equipment return without him.
The System had awarded his contribution because a model survived verification. The desk offered a life built around making more such models. Refusing it would not mean the work was beneath him. It would mean the distance from evidence to action mattered enough that he was willing to pay for proximity.
An enemy could be resisted. A sensible future had to be declined with full knowledge of what it offered.