The Long Report

The inquiry sat for the first time nine months after they left, in a room in Palun with eleven people in it and a live feed to four hundred more, and the first witness was Ivo Marchetti, because he was the assessor of record and because Rey had refused.

He was on the stand for two days. He had prepared, in his own way, which was to write out every answer he thought he might have to give in the small crabbed hand of a man trained on paper, and then to leave the pages in his cabin, on purpose, and answer without them.

The transcript is a document that engineers still read. Not for the technical content, which is better set out in volume four, but for a four-minute exchange on the second afternoon in which counsel put it to him that the maintenance interval had been irresponsibly long, and Marchetti agreed, and counsel moved on, and Marchetti said, "No, I would like to finish, if I may."

What he said was that a twenty-year interval is irresponsible only if you assume the world lasts twenty years, and that the engineers who set it had assumed a company would last twenty years, and that this had been, at the time, an entirely reasonable assumption held by every person in the industry including everybody presently in the room.

Then he said the thing that got quoted. He said that the failure at Hesper was not that a machine had been left alone, but that a promise had been made by an organisation to a machine, and that organisations cannot keep promises across ninety years because they are not the kind of thing that lasts, and that the machine had kept its half of it to the end and rather beyond.

Counsel asked him whether he was suggesting the foundry had been wronged. Marchetti said no, and then sat and thought about it for eleven seconds with four hundred people watching, and then said: "I am suggesting that if a thing does exactly what we asked for ninety years and we do not come back, the word for what went wrong is not a technical word, and I have not got a better one."

Sabine watched the feed in a workshop on deck nine with her hands over her mouth. Rey watched it in a bar and had to leave. Marchetti came off the stand, apologised to both of them separately for having been sentimental, and was extremely annoyed for about a year that it was the only part anybody ever mentioned.

Sabine Tarr's paper was published fourteen months after they left and it is nine pages long, and eleven thousand people have read it, and it is not about lonely machines.

It is titled, in full, Recovery of ninety years of process history from tool-path residue on interior plating at an unattended orbital foundry, and it is written in the flattest prose she could manage, on purpose, because Rey had told her in the last week on the station that the surest way to have this taken seriously was to write it as though it were dull.

The method section is the part that mattered. It sets out, step by step, how to get from a surface scan to a campaign record, using a documented tool-path generator, with worked examples, and it does so in a way that generalises to any facility whose dressing routine is thermally driven, which is most of them.

In the eleven years after publication the same method was run on nine other unattended facilities, and it found nothing at all at six of them, and at three of them it found drift that nobody had known about, one of which was in the other direction and did matter, urgently, and was corrected four months before it would have become a very different kind of report.

Sabine's acknowledgements name Ochoa, who died in the year the last crew left plus eleven, and whose appendix on page nine hundred and six she reproduces in full as an appendix of her own, on the grounds, stated in a footnote, that it is the actual method and that she had merely noticed it was still there.

She was asked once, at a conference, whether the walls were beautiful. She said yes without hesitating, and the room laughed, and she waited for it to stop and then said that she wanted to be clear she was not being charming: that they were beautiful, that this was an accident, that the accident was worth about four hundred million in avoided survey costs, and that she would like the field to be considerably less embarrassed about noticing when a thing is beautiful, because she had personally wasted nine days being embarrassed about it and had found nothing during those nine days at all.

That is also quoted, rather more often than the method section, which she has made her peace with. She minds it a great deal less than Marchetti minds his, and has said so to him twice, both times at dinners neither of them wanted to attend, and both times he told her that this was because she was thirty-four and would learn.

Rey Castellan was offered a book contract, twice, and a documentary, and a post at the institute, and a promotion to a desk in Palun that would have doubled her money, and she took none of them and went back out on a survey contract eleven months later with three people she had never met.

She was asked about it, in the way people are asked, at the sort of dinner she attended about once a year and disliked. She said that she had spent ninety-one days in an empty building and had come away with one thing she actually believed, and that it did not fill a book.

The thing was this: that the melt hall was the most complete and honest record she had ever seen of anything, and that it existed because a maintenance routine had been left running with nobody to switch it off. Every deliberate record she had made in nineteen years, every log, every handover, every carefully written report, had been shaped by knowing that somebody would read it. The walls had not been.

She did not think this made the walls better than a log. She thought it made them different in a way she could not get to the bottom of, and she had tried, in the empty crew quarters at hours that meant nothing, for eleven weeks, and had failed, and had eventually concluded that failing to get to the bottom of it was a reasonable thing to spend a decade on and rather a poor thing to write nine chapters about.

What she did instead was change how she ran crews. Small things. She stopped taking the amber alarms off the summary panel. She made everybody read the consumables list out loud once a week, which was mocked for about four contracts and is now, in a diluted form, in the standard operating procedure of two hundred vessels.

And on every station she ever worked afterwards, at some point in the first week, she would find whatever the oldest continuously running process on it was, and go and look at it, and read its log back as far as it went, on her own time, for no operational reason at all. She never once found anything. She kept doing it for twenty-two years.

She was asked about that too, eventually, by a young systems engineer on a hauler, who had heard the story and wanted to know what she expected to find. Rey said she did not expect to find anything. The engineer said then why do it, and Rey said, "Because it is going one way or another whether I look at it, and it costs me an afternoon," and went back to what she was doing.

Eleven days after they undocked, at nineteen hundred and four by a clock that had not been set by anybody since the last crew left and had drifted eleven minutes in ninety years, Hesper Foundry began campaign four thousand two hundred and twelve. There was no decision in it. The order book said repeat and the schedule said nine days and the nine days were up.

It brought the melt hall up over nine hours, drew feedstock from the number three hopper, ran the roll for eleven and a half hours, cut and stamped eleven-metre lengths and racked them, measured every one against the rolling mean of the last two hundred accepted, passed all of them, adjusted the gap by four ten-thousandths of a millimetre, and printed a certificate for each, which it clipped to the rack.

Then it generated a maintenance report, four pages, giving among other things the cumulative divergence from specification to four decimal places and a recommendation that the roll be re-baselined at the earliest opportunity, and sent it to a printer on deck two that had no medium in it, and raised a consumable alarm that was already raised, and displayed it in amber on a panel eleven centimetres wide in an empty corridor where three people had walked past it, twice, eleven weeks earlier, on their way somewhere else.

Then it ran the surfacing head over the melt hall plating, along a path computed from the campaign's thermal map, starting one fingernail's width beyond where it had finished the last time, and cut a band about a hand's width across into steel that had been cut into four thousand two hundred and eleven times before.

Then it dropped the hall lights section by section on the standard occupancy logic, put its long face to the sun, trimmed with a burn of eleven milliseconds out of a propellant budget it had been managing carefully for ninety years, held its attitude to a tenth of a degree, and waited nine days for the next order in a book that had last been updated forty years earlier by a scheduling clerk in an office that no longer exists, and that instructed it, on the last line, to repeat.

There was nobody aboard. There had been nobody aboard for eleven days and there would be nobody aboard for another forty-one years, and the coolant loop made the sound it makes, and the figure on the wall went one more step further from where everything had started, and none of this was tragic, and none of it was lonely, and it did not know, and it did not stop.

Somewhere on deck two, taped inside a panel, a card in block capitals went on saying DO NOT ADJUST THE GAP WITHOUT READING P411, in the dark, patiently, to a corridor, the way it had for ninety years and would for forty more, until a woman of sixty-eight came back with nineteen people and a mandate, and read it, and stood there for a while, and then went and found the page.