The batch failed on a Wednesday.

The fourth neutralizer batch -- the one he had been running in parallel with the stabilization protocol, building toward the extended duration window that the restorative mechanism was capable of -- came back from the overnight synthesis with a contamination marker in the cascade inhibitor layer. Not catastrophic: the contamination was caught before the compound was used. But the batch was lost. Three weeks of parallel work, gone.

He told her in the morning.

She read his face before he spoke, which she always did, and he had stopped being surprised by this and had started feeling something else -- something that he did not always examine directly, which was the specific quality of being known accurately by another person. She read his face and said: "Batch."

"Yes."

"Mechanism?"

He walked her through the contamination: the cascade inhibitor layer's chemical stability had degraded at a specific temperature junction in the overnight process, which meant the synthesis protocol needed an adjustment to the temperature profile, which meant rebuilding the batch from the inhibitor synthesis stage. Significant work. Not impossible work.

She listened. She asked two questions -- both technical, both to the point -- and the second one surfaced something he had not seen: a timing consideration in the inhibitor layer's synthesis that he had built into the protocol based on the original batch framework but that the new stage one mechanism might require him to reconsider from scratch.

He looked at this.

"You're right," he said. "The timing is wrong for the stage one intervention point. The original framework was designed for stage three targeting. I carried the timing forward without adjusting for the new mechanism."

"It's a significant adjustment," she said.

"Yes."

"How long?"

He thought about it. "Six to eight days to redesign the inhibitor synthesis profile. Then another full batch cycle." He looked at the failed batch documentation. "The new batch will be different from the previous ones. I don't want to run it without rebuilding the protocol from the mechanism level."

She nodded. She went to the analysis room. He went to the research wing.

He rebuilt the protocol.

---

In the middle of the rebuild, he returned to the alternative pathway analysis.

He had been running this in the background for eleven days -- whenever the primary synthesis work allowed, he was looking at the alternative. He had been systematic: working outward from the original model's assumptions, checking each assumption against the available literature, looking for the gap that would change the conclusion. He had found gaps in three assumptions. Correcting all three had not changed the conclusion.

He was now at the boundary of his existing knowledge. The alternative pathway, if it existed, would require a mechanism he had not yet considered. He needed a different perspective.

He thought about who could provide it.

Anwen could not -- she was a patient with research-adjacent knowledge, not a synthesis specialist. Bren was logistics, not biochemistry. The Bellhaven research registry had been consulted indirectly through the provisional filing, but he had not requested a direct research consultation because a direct consultation would require explaining what he was looking for, which would require explaining why.

There was one other option.

He had been aware of this option since he had first found the Heliogen records fragments with the erasure signature. He had not acted on it because acting on it would require finding the person whose signature was on those records, which had meant going through channels he did not have direct access to. He had asked Bren to look. Bren had been looking for two weeks.

He messaged Bren through the relay: *Any progress on the contact search?*

Bren's reply came in thirty minutes: *Working on it. Northeast network has leads but unconfirmed. Give me a week.*

He acknowledged. He went back to the protocol rebuild.

---

The new batch synthesis began on the fifth day of the rebuild.

He ran the modified inhibitor synthesis profile first -- testing the timing adjustment in isolation before committing to the full batch. The test ran overnight. In the morning the test result was clean: the cascade inhibitor layer was stable at the new temperature profile. He began the full batch.

Mirelle came into the research wing to see the result of the test. She read the monitoring display. She said: "Clean."

"Yes."

"The timing adjustment worked."

"The test suggests it will. The full batch is running now." He looked at the display. "Three days to completion."

She looked at the monitoring data. She was developing an accurate read of synthesis monitoring, which was a thing he had not expected and which had been arriving gradually since the mechanism analysis work -- the specific competence of someone who had been surrounded by this information long enough to start making reliable predictions about it.

"The cascade inhibitor layer looks different from the previous batches," she said. "The stabilization curve."

"Yes. The stage one targeting changes the inhibitor's behavior in the cascade. The new mechanism produces a steeper initial stabilization and a more sustained plateau." He looked at the curve. "The new batch should have a longer effective window."

She looked at this. "Longer than the last batch?"

"Possibly significantly longer. We won't know the specific duration until we test."

She was quiet for a moment.

He was aware of the specific quality of her silence -- she was thinking about the testing protocol, about what the new batch would mean in terms of session, about the duration window. He was aware that this was a specific kind of awareness that he would not have had six months ago.

He said: "When the batch completes, I'll want to run a safety assessment on the new inhibitor profile before we do any contact testing. Two days minimum."

"All right."

"And I want to introduce the new compound gradually -- shorter session first, confirm stability before extending."

She looked at him. "You're being careful."

"I'm always careful."

"I know. I'm noting it."

She went back to the analysis room.

He continued monitoring the batch. At midday he went to the kitchen for lunch and found her at the table with a letter from Anwen -- the second one that week, which meant Anwen was writing more frequently, which he had noticed and which he had been noting as its own form of data: a person who is waiting for treatment has more to say.

He sat at his end of the table and made lunch. She read the letter. He watched her read it -- the specific quality of her face when she was reading Anwen's handwriting, which he had been able to distinguish for months now from the face she made reading relay documentation or transit maps.

"She found the second book," Mirelle said.

"Already?"

"She had Calla find it." She looked up from the letter. "She says Calla is threatening to read the series herself so Anwen will have someone to discuss it with."

He thought about Calla, who he had met twice at the hospital and who had the specific competence of someone who had been caring for long-term patients long enough to know that the emotional work of caring was as important as the medical work. "That sounds like Calla."

"Yes." Mirelle set the letter down. "She says the light in the corridor is different in the late afternoon now. The angle has shifted with the season." She paused. "She says she's been sitting in it at a different time to follow the change."

He thought about Anwen tracking the transit hour equivalent in a hospital corridor, adjusting her schedule to follow the light as it moved through the season. He thought about the specific persistence of it -- not just finding the light once and sitting in it, but following it as it changed.

"She's going to like the greenhouse," Mirelle said.

He looked at her.

"When she comes to visit," she said. "She's going to like the greenhouse."

He held this. He thought about Anwen in the greenhouse, standing in the UV-filtered light without a mask, receiving the morning the way Mirelle had described. He thought about Anwen seeing the hydroponic trees flower and coming to the northeast perimeter sector and sitting in the light room with the stone floor.

He thought about all of this and about the encrypted folder and about three days until the batch completed.

He said: "She will."

They finished lunch. They went back to their respective work. The batch ran its cycle in the lab and the transit hour came and went and the afternoon proceeded.

At three o'clock the monitoring system flagged a minor anomaly in the inhibitor stabilization -- not the contamination from the failed batch, something different. He ran the diagnostic. The anomaly was in the external contact layer of the compound.

He ran the diagnostic three times. The result was consistent. The external contact layer, which was the layer designed to interface with biological material at the point of application, had a reactivity profile that was slightly elevated from the expected range.

This was the layer that would contact Mirelle's skin when the batch was tested.

He ran additional diagnostics. He examined the elevated reactivity from several angles. He was not certain it was a problem -- it was within a range that could still be acceptable, depending on the specific sensitivity parameters. But it was elevated, and he did not know why.

He noted it. He continued monitoring. He would know more when the batch completed.

---

The batch completed on the third day.

He ran full characterization before doing anything else. The compound was viable: the stage one mechanism was functional, the inhibitor layer was stable, the duration window appeared significantly extended. The external contact layer reactivity was still elevated -- confirmed across all characterization assays.

He told her about the batch completion that morning.

He also told her about the reactivity.

"The external contact layer is showing elevated reactivity," he said. "I don't know the significance yet. It's within a range that might be acceptable but I'm not prepared to say it is until I've run more analysis."

She looked at him. "What does elevated reactivity mean in practical terms?"

"It means the layer may interact with biological tissue more actively than the previous batches. In the best case, it's a neutral effect -- the compound works as intended without side effects. In a worse case, it could cause a contact response at the application site." He paused. "I need more time to characterize it before we test."

She nodded. "Tell me when you know."

He told her four days later.

The contact layer reactivity was not neutral. It was not catastrophic -- not a reason to discard the batch -- but it was a real effect: a chemical burn response at the application site, localized, manageable, painful. He had confirmed this through a series of controlled tissue exposure tests using cell cultures from the station's basic biological stores.

He sat with this for a day before telling her.

He told her on the fifth day after the batch completion.

She listened. She asked the specific questions she always asked: severity, duration, what it meant for the treatment protocol if the reactivity was consistent across future batches.

He answered. He told her: severity was moderate, duration typically twelve to twenty-four hours for full resolution, and no, the reactivity profile should be correctable in the next synthesis iteration -- it was a formulation issue, not a mechanism issue.

She nodded at each answer. When he had finished she said: "Show me the tissue exposure results."

He showed her.

She looked at the cell culture data for a long time. He watched her read it with the specific attention of someone who understood what the data meant, not theoretically but through the specific knowledge of what exposure to chemicals felt like on a body that had been exposed to chemicals designed for her body specifically for twenty-three years.

She set the data down. "All right," she said. "Fix the formulation."

"It will take another batch cycle. Another three weeks."

"I know." She looked at him. "The timeline for Bellhaven -- does this affect it?"

"Not substantially. The stabilization compound is already in transit." He paused. "Anwen's treatment uses the stabilized compound, not the neutralizer batch. The neutralizer is separate."

"I know that too." She stood. "Fix the formulation. Let me know when the new batch is ready."

She went to the Surface Room.

He sat in the research wing with the batch characterization data and the tissue exposure results and the encrypted folder and the alternative pathway analysis that had not produced a result and Bren's message from three days ago that said *one more week, I have a lead.*

He thought: fix the formulation. Find the alternative. Find Ivara Pell.

He had been at the station for one hundred and sixty-one days.

He went back to work.

---

The actual exposure happened on Day 163.

He had been rebuilding the formulation -- the fifth day of that process -- and she had been running the route extension plan, and at three in the afternoon she came into the research wing. She was not bleeding but she was holding her left forearm in the specific way of someone who had sustained a contact injury.

He looked at her arm. "What happened?"

"The batch vial in the supply cabinet," she said. "I was moving storage to make room for the relay equipment. The seal was compromised. Contact, maybe four seconds."

He was already at the supply cabinet. The vial he had stored there after the characterization -- not the active batch, the secondary characterization sample. He should have labeled it more clearly. He should have moved it before she had access to the supply cabinet. He had not.

He took her to the examination room.

The burn was what the tissue exposure data had predicted: a contact response at the application point, covering roughly the size of his palm on her inner forearm. It was in the early stage -- the reactive phase, before it settled into the resolution profile. It looked worse than it would be in twenty-four hours. This did not make it less painful right now.

He cleaned it. He applied the topical treatment he had prepared for exactly this eventuality -- the same eventuality he had tested for and confirmed as likely and then stored the characterization sample inadequately.

She held still through the cleaning. She was very controlled about physical discomfort -- he had observed this in the wound closure after the transit hub fall, the same specific quality of someone who had learned to manage pain through high tolerance and deliberate stillness.

"This is my fault," he said.

She looked at him. "The seal was compromised."

"I should have moved the sample before it was in the same storage space as the general supplies. I was aware of the reactivity profile." He applied the topical. "I was focused on the formulation rebuild and I didn't complete the storage reorganization I had planned."

She said: "You've been focused."

"Yes."

"On the formulation rebuild."

"Yes."

She was looking at him with the assessment quality. Not hostile -- reading. He was aware of it and he was aware of what she was reading: whether his focus on the formulation rebuild was the only thing occupying him, or whether there was something else.

He met her eyes. He held the look.

"The formulation is the primary problem right now," he said. "That's where my attention is."

"I know," she said.

He finished the treatment. He bandaged the burn. He went through the post-treatment instructions -- the same precision he brought to the wound closure protocol -- and she listened with the same focused attention she brought to information she was going to retain and use correctly.

When he finished she looked at the bandage.

"How long?" she said.

"Twenty-four hours for most of the acute phase. Full resolution in three to four days."

She nodded. She stood.

"Mirelle," he said.

She looked at him.

"I'm sorry," he said. "For the storage failure. That was mine."

She held his eyes. "I know it was."

She went to the analysis room.

He sat in the examination room for a while after she left. He thought about the tissue exposure data and about the storage failure and about the specific word careless, which was not a word he normally associated with himself and which was the accurate word for what he had done. He had known the reactivity profile. He had known the storage was inadequate. He had been focused on other things.

He thought about what the other things were.

He was focused on the formulation rebuild. He was also focused, in the background of everything, on the encrypted folder and the alternative pathway analysis and Bren's lead that had now been a week in progress. He was running multiple things in parallel and one of them had slipped and it had produced a burn on her forearm.

He stood up. He went back to the research wing.

He ran the storage inventory and relocated all characterization samples to secure containment, properly labeled, in a location she would not be accessing as part of standard station maintenance. He did this completely and correctly.

Then he went to the synthesis files and worked on the formulation rebuild.

The next day, she brought him tea at the workstation. She did not mention the burn.

He did not mention the encrypted folder.

They worked in the standard parallel pattern, and the formulation rebuild progressed, and the day proceeded.

He thought: this is still manageable. The burn was a failure of attention. I will not let my attention fail again. I will complete the formulation rebuild and I will find the alternative pathway and I will not let anything else slip.

He thought about this and believed it and did not examine whether the believing was the same as the truth.