The Fourteen-Day Window

The Sentinel Arc, Book Two — Part Two: The Moon

by N.W.

1. Shackleton, 2061

The station was built into the side of a crater rim and from the outside it looked like a mistake — a grey box on grey ground under a black sky, with a comms mast that leaned four degrees because the regolith under one footing had compacted differently and nobody had thought it worth the argument.

Nadia had been there fourteen months.

She was not an astronaut. That word had cost her three years and she had stopped using it around the time she understood that it described a job nobody at Shackleton actually did. What she was, on the roster, was Medical Systems — one of two, on a station of nineteen, which meant that when the other one slept she was the whole of it.

The assessment that had let her come was not the one she fought in Solna. It was a different envelope: eleven light-seconds to Earth, a cycler on a regular arc, and a return window every fourteen days. Against that envelope her twin had returned 6.1 per cent, which is under eight, and the determination had taken four days and nobody had asked her a single question about it.

She had wanted to feel vindicated. What she felt was that the six years had been spent arguing about the wrong number, and that nobody had told her because nobody had known.

Fourteen months does a specific thing to a person. The first two are photographs. The next four are work. After that the place stops being anywhere in particular and becomes the ordinary background against which you are tired or not tired, and Earth becomes a light in the window that you notice about as often as people at home notice the Moon.

She had expected to think about the Moon constantly. What she thought about was the water reclamation cycle, whether Bakker was sleeping, and a running argument with herself about whether she was any good at this.

The nineteen were: six on regolith and extraction, four on power, three on habitat and life support, two on comms and computation, two medical, one commander, and Bakker, who was a geologist and the only person on station doing what he had trained for.

2. Ovechkin

The first she knew of it was the sound of somebody being too polite.

"Sorry — is this a bad time?"

Yuri Ovechkin was forty-four and did power systems and had the specific over-courtesy of a man who has decided not to mention something and then mentioned it anyway. He was standing in the hatch with one hand on the frame.

"It's never a bad time. Sit."

"It's nothing."

"Then it'll take a minute."

He sat. He had cut his hand nine days ago on a panel edge in the regolith processor, cleaned it himself, logged it because the log is mandatory, and thought about it exactly as much as anyone thinks about a cut hand. It had closed. There was no redness worth the word.

"I'm tired," he said. "That's all. I've been tired since Thursday."

Nadia looked at him for a while.

"Anything else."

"No."

"Yuri."

"I was cold last night," he said. "In bed. Which is stupid, because it was twenty-one degrees."

She had known him for fourteen months in the way you know someone you eat two meals a day with and have never once discussed anything of consequence with. He had a daughter at university in Kazan, he was better at chess than he admitted, and he had the habit — common among people who fix things — of describing a failure in terms of what he had already ruled out.

"I ruled out the obvious," he said, unprompted. "I slept badly, and there was the thing with the regulator on Thursday, and I am forty-four, which I am told is a diagnosis on its own."

"None of those makes a person cold at twenty-one degrees."

"No," he agreed, comfortably, as though they were discussing somebody else's power system.

3. What the model said

She pulled his twin at 19:40 and read it standing up.

It had been fitted when he was twenty-nine, at a hospital in Nizhny Novgorod, for a reason unrelated to anything — a cardiac question that had resolved into nothing. Fifteen years of refits. A good model, well-behaved, boring in the way you want.

At 19:44 it flagged.

Not a diagnosis. Models do not do diagnoses; they do trajectories. What it said was that his heart rate variability, his overnight temperature curve and the shape of his last two days of activity were consistent with an inflammatory trajectory that, projected forward under the station's conditions, reached organ dysfunction in somewhere between forty and ninety hours.

Probability, over the envelope: 31 per cent.

Nadia read the number and felt the floor of her stomach go.

Then she did the thing she had spent six years learning to do, which was to open the credibility file.

4. Page twenty-nine

It was on page thirty-one for his model, but it was the same row.

Sensitivity to initialisation prior: 0.7 percentage points. Sensitivity to environmental envelope: 11.4 percentage points.

She sat down.

The first number was the one that had nearly ended her life plan, and on Ovechkin's model it was nothing — his fit had begun on an ordinary Tuesday about an ordinary question, and fifteen years of ordinary Tuesdays had washed it out entirely.

The second number was one she had never had to think about, because on Earth it is always small. Environment is the thing a model knows best. Air, pressure, gravity, the ordinary constants.

Except here they were not constants. They were settings. The station ran at 0.16 g and at a partial pressure that a committee had chosen, in an atmosphere recirculated through filters whose performance curve was itself modelled rather than measured, and Ovechkin's twin had been fitted across fifteen years in which none of that had ever been true.

Eleven point four points of that thirty-one came from a machine reasoning about a place it had never seen.

5. The window

She went to find Adeyemi, who commanded, and who was eating.

"How long to the window."

He put the fork down. "Which window."

"Return."

"Six days." He was already standing. "Who."

"Ovechkin. Maybe."

"Maybe."

"Thirty-one per cent over the envelope. Eleven of it is the model guessing about the environment." She heard herself and slowed down. "Sorry. Thirty-one per cent probability of organ dysfunction inside ninety hours. The estimate is soft in a specific way and I can tell you exactly how soft."

Adeyemi was quiet for a moment.

He had been at Shackleton for two rotations and had put one person on a cycler in that time. It had been a psychiatric evacuation, and it had been correct, and he had spent a year being asked about it in the tone people use when they have already decided. The board had cleared him. The question had kept arriving anyway, in slightly different words, from slightly different people, for eleven months.

What that had taught him — and he had said this to Nadia once, in the galley, and never again — was that the review of a decision to evacuate is conducted by people who know how it turned out, and the decision itself is made by a person who does not.

"If you send him," he said, "you send him on the six-day window and he is on that vehicle for another nine. Fifteen days in a can with a medical kit and no you."

"I know."

"And if you don't and it's real, then at hour ninety he is here, and here is eleven light-seconds from a consultant and six days from a bed."

"I know that too."

"So what do you want."

And there it was — the thing she had asked a room in Solna for, six years ago, and got.

"I want it to not be a number," Nadia said.

6. What she had asked for

Here is what the disclosure reform had actually produced, four years after a determination was vacated in a room in Solna.

Every credibility file now carried its sensitivities on the front page. Every estimate arrived with the statement of where it was weakest. It had been implemented well, by people who understood it, and it had cost about eleven million euro across three agencies and had been called a model of proportionate regulation in a paper Nadia had not read.

And what it meant, at 20:15 on a Tuesday at Shackleton, was that a thirty-one-year-old with one colleague asleep and eleven light-seconds to anyone else was holding a number and a statement of the number's weakness, and both of them were true, and neither of them told her what to do.

She had won the argument. The argument had been that the doubt should travel with the number.

Nobody, including her, had thought about who would be standing there when it arrived.

7. The ground

She sent the packet at 20:22. Eleven seconds up, some minutes of a person reading, eleven seconds back.

The reply came from a flight surgeon in Cologne called Weiss, who she had spoken to twice and liked.

Seen. Concur the trajectory is plausible. Concur the envelope sensitivity is material. I can't resolve it from here — the model that would resolve it is the one you're asking about.

What do you have that it doesn't?

Nadia read that four times.

Then she went and looked at Ovechkin.

8. Twenty-one degrees

He was in the galley making tea he did not want, because that is what people do when they have been told to stay where someone can see them.

"You said you were cold," Nadia said. "In bed. At twenty-one degrees."

"It's nothing."

"When."

"Sunday. And last night." He shrugged. "It passed."

"Did you shiver."

He thought about it, and she watched him think about it, and she saw the moment he decided to tell the truth rather than the polite version.

"Yes," he said. "For about ten minutes. Both times. My teeth went."

She wrote it down. Not into the model — into a notebook, with a pen, because the model had no field for it.

Rigors. Twice. Nine days after a breach of the skin.

The twin had every number: temperature, variability, activity, sleep. It did not have the fact that a man had lain in the dark with his teeth chattering at twenty-one degrees and had decided this was not worth mentioning to anyone, because he was tired and it passed and he did not want to be the person who made a fuss.

That is not a signal a body emits. It is a thing a person tells you, if you ask them the right way, standing up, in a galley, at nine at night.

9. What she sent

Two rigors, Sunday and Monday, ~10 min each, teeth chattering, ambient 21C. Nine days post skin breach, regolith processor panel, cleaned and closed, logged 12/7. Not in the model. He did not consider it reportable.

Weiss came back in under four minutes, which meant she had been sitting there.

That changes it. Not the number — the number is the number. It changes what I think the number is about.

Recommend you treat as presumed bacteraemia and put him on the window. My concurrence is on the record and it is based on your observation, not on the model.

Nadia — write down that last sentence somewhere. You will need it.

She did not understand the last line for about four months.

What Weiss had done, in one sentence, was move the basis of a clinical decision off an instrument and onto a person, in writing, on a record that would be read by a review board if it went badly. It is an ordinary sentence and it is also the whole of what a senior clinician can give a junior one at eleven light-seconds: not the answer, but the willingness to be named alongside it.

Nadia has since watched two people fail to do that, in circumstances where it cost nothing, and she has stopped assuming it is ordinary.

10. Six days

They did not have six days of doing nothing. They had six days of Ovechkin on broad-spectrum antibiotics from a formulary of eleven, chosen against a pharmacokinetic model whose own credibility file admitted that nobody has ever measured beta-lactam clearance at a sixth of a gravity in a human being.

That number, she noticed, was not on the front page of anything. The reform had covered diagnostic estimates. Dosing models had been a separate workstream and had been deferred at the 2058 review for reasons that were, when she looked them up later, entirely reasonable and had to do with committee capacity.

The six days were not dramatic and she has been careful, since, to say so, because the version where they were dramatic is the version people want and it teaches nothing.

He worked, at reduced hours, because there was no medical reason to stop him and every psychological reason not to. He ate. He beat Bakker at chess twice and lost once, which Bakker reported to her unasked, in the manner of a man delivering a clinical observation. She saw him four times a day and wrote something in the notebook each time, mostly the word "same".

Bakker, who was not medical and had no standing to say anything, stopped her in the corridor on day three and said: "He is walking differently."

She had not noticed. She went and looked. He was walking differently — very slightly, favouring nothing in particular, the gait of a person conserving something. It went into the notebook and into the packet to Cologne, where Weiss read it and replied with one word: noted.

He improved on day two, plateaued on day three, and on day four his lactate — which they could measure, badly, on a point-of-care unit certified for a range that did not quite cover what they needed — came back at a value she looked at for a long time.

Not high enough to act on. Not low enough to stop thinking about. The unit's own documentation described its accuracy at that end of the range as indicative, which is a word that means the instrument would like you to stop asking.

On day six he walked to the vehicle himself, which he insisted on, and which Adeyemi allowed because there is a difference between a patient and a man being carried in front of his colleagues.

11. Nizhny Novgorod, eleven weeks later

They cultured him on Earth and got Staphylococcus aureus, methicillin-sensitive, from a set drawn on arrival. He had been on the right drug for fifteen days by then. He recovered, fully, and went back to work — not to the Moon, because the reassessment after a septic episode is what it is, and he has been philosophical about that in a way Nadia finds harder to look at than bitterness would have been.

The interesting part is what the review found.

She was interviewed twice, by video, eleven light-minutes of nothing between question and answer the first time because the ground scheduler had booked the wrong relay. The second was better. A woman called Fournier asked good questions for two hours and one bad one at the end, which was whether Nadia had considered that she might have been pattern-matching to her own history.

She had. She had considered it at 20:04 on the Tuesday, before she opened the credibility file, and had written it in the notebook: am I doing this because of me. The page is still there.

The honest answer is that she does not know and cannot know, and that the reason it did not matter is that the thing which changed the decision was not a pattern. It was a fact she went and got.

The twin had been right. Thirty-one per cent had been, if anything, conservative; the retrospective fit put it nearer forty. The environmental sensitivity had been real and had been in the direction nobody expected — the model had underestimated the trajectory because it assumed a rate of immune clearance that does not hold at 0.16 g.

And the reason Ovechkin lived was a note in a paper notebook about a man being cold at twenty-one degrees.

The review recorded that. It recorded it carefully and at length and recommended that rigors be added as a structured field in the crew health record, which was implemented in eight months and is now a checkbox.

What it did not record, because a review has no field for it either, is that the checkbox exists because somebody stood in a galley and asked a second question after the first one had already been answered politely.

There is a version of this account in which that is the moral, and it is a bad version. Asking a second question is not a method. It is what a person does when they have the time, the standing and the specific suspicion — and Nadia had all three only because she had spent six years of her life reading a document type that almost nobody reads.

The next medic will not have had those six years. That is the actual problem, and a checkbox does not touch it.

12. Bakker

She told Bakker about it in March, four months later, because he asked and because he was the one who had noticed the gait.

He listened the whole way through, which is rarer than it sounds, and then said: "So the machine was right."

"Yes."

"And you nearly didn't send him."

"I nearly didn't send him for six hours. Then I did."

"Because of the shivering."

"Because I asked and he answered."

Bakker turned that over. He was a geologist and had spent his professional life inferring enormous things from small ones, and Nadia has thought since that this is why he got to the point faster than the review board did.

"Then the machine was right and useless," he said. "At the same time. It told you thirty-one per cent and thirty-one per cent is exactly the number where you cannot do anything with it."

"That's not fair. It told me to look."

"It told you to look," Bakker agreed. "Which is different from telling you what to do, and everybody keeps saying those are the same thing."

13. The thing nobody was assessed against

Nadia wrote to Halvorsen that winter. It was the first time in four years.

She described Ovechkin, and the credibility file, and the eleven point four, and Weiss's sentence about the concurrence resting on the observation rather than the model. She wrote it badly, in three drafts, and sent the third.

Halvorsen wrote back in a week, which for her was fast.

You have found the next one.

The reform put the doubt on the front page and I still think that was right. What it did not do — what nobody asked it to do, including you, including me — was work out what a person should DO with a number that comes with a stated weakness. We built an instrument that tells the truth and we handed it to people with no training in what a truthful instrument feels like when it is uncertain.

You had six years to get ready for that moment. Ovechkin's colleague at the next station will have had a two-day course.

I do not know what the answer is. I know it is not another disclosure requirement, because we have already done that one and you are the proof that it worked.

Write it up. Not the case — the gap. You are the only person who has stood on both sides of it.

14. What she has not written

She has started it four times.

The difficulty is not the argument. The argument is clear enough: an estimate that declares its own weakness transfers the judgement to whoever is holding it, and transferring judgement is not the same as removing burden — it is the opposite, and the people it lands on are the ones with the fewest colleagues and the least time.

The difficulty is that she cannot write it without describing the alternative, and the alternative is a fixed threshold applied by a tired person at four in the morning, which is the thing she spent six years arguing against, and which — she now understands, in a way she could not have understood at seventeen — Ekwueme had been entirely right about.

Both are true. That is what makes it a hard problem rather than a scandal.

She keeps the notebook. It is a real notebook, paper, and it has one useful line in it and about forty pages of nothing, and it goes wherever she goes.

15. The window, again

Shackleton has a return window every fourteen days.

At the surface stations that is a constraint you plan around: you know it, you feel it, and on day eleven of a cycle everyone is slightly more careful than they were on day two. It is the outer edge of what can go wrong before something becomes irreversible, and it is short enough that a person can hold it in their head.

The transfer window to Mars opens for about three weeks every twenty-six months.

Nadia has been reading about that for a while now — not as a plan, as a shape. On the surface of the Moon a mistake costs you six days and a cramped vehicle. Out there it costs you everything, because there is no window, only a departure and an arrival separated by an interval during which whatever is true has to stay true.

Every argument in this account gets harder there. The environmental sensitivity is worse, because nothing about the envelope has ever been observed in a human. The dosing models are worse, for the same reason. The disclosure reform helps less, because there is nobody eleven seconds away to send the packet to.

And the thing that saved Ovechkin — a person asking a second question after the first was answered — is the only part that scales, because it costs nothing and requires no instrument and works at any distance from Earth.

That is not a comfortable conclusion for anyone in this story to have reached, including the woman who fought a model and won.

Part Three: The Twenty-Six Month Interval — Mars.

Fiction. The clinical and modelling substance — post-septic reassessment, environmental sensitivity in a fitted individual model, the absence of measured pharmacokinetics in partial gravity, and the gap between an estimate that states its own uncertainty and a person who has to act on it — is drawn from the published literature set out in The Waiting Problem (WINNIIO, 2026), where the sources are given. Rigors following a skin breach are a real presenting sign and a real omission from structured records. The Moon here is a place where two things are true at once, which is the only way it resembles the one we will actually build.