Budget Line 9 Fell 61 Per Cent. The Sightings Changed Shape Six Weeks Later.

GREEN BANK, W.Va. — Sixty-three per cent of the sighting reports fall inside ten per cent of the day.

Not our day. The observing post's day, which is thirty hours long and drifts against ours, and which is the reason nobody had noticed.

Line 9 of the Committee's Peripheral Observation vote pays for the watch on this system. At the 2019 review it was cut by 61 per cent. The post went from three continuous watches to one, staffed fourteen hours in a thirty-hour local day, and the work was let to a contractor.

Six weeks after the contractor took over, the shape of terrestrial sighting reports changed. It has not changed back.

The two windows

A single watch cannot hand over to itself.

So twice a local day, for ninety minutes, the post runs both instruments at once: the outgoing watch keeps its own instrument up while the incoming watch brings the second one to temperature and takes the briefing.

Three hours out of thirty. Ten per cent of the local day. In our 214 reader reports PX-2041, those three hours hold 134.

Before 2019 there were three continuous watches and no overlap at all, because nothing needed handing over across a gap.

Why nobody saw it

Because of the drift.

The post's day is thirty hours and ours is twenty-four, so the handover windows walk against the terrestrial clock — about six hours per local day, which carries them right round our twenty-four-hour dial every four local days, or five of ours.

That means the reports do not cluster at ten in the evening, or at any other time you could name. They cluster at a time that moves. On any ordinary sort-by-hour, a cluster that walks looks exactly like no cluster at all.

Every published analysis of this dataset sorted by terrestrial hour and found nothing. That includes two of our own, in 2021 and 2023. Both said there was no pattern. Both were wrong, and the reason they were wrong is that we sorted the file by our own clock and never asked whose clock the sky was on.

The test, run blind

We gave Dr. Onome Halvorsen-Baptiste 214 timestamps and no context.

She was asked one question: over candidate day-lengths from twenty to forty hours in steps of a tenth of an hour, is there any period at which these reports clump? Her analysis plan was registered before she saw the data.

She came back with 29.9 hours and two windows about ninety minutes wide.

Then we showed her the roster. The windows matched to within eleven minutes.

The operations manager

Ilma Voskuijl-Terry has run operations for the contractor since the contract was let, and answered the question the way somebody answers it for the third time.

"We run both instruments across a handover because the outgoing watch has to hand over something, and you cannot hand over a sky you have stopped looking at," she said. "It is ninety minutes. It is in the roster. The roster is published, because publishing it is a condition of the contract, and nobody reads it."

She has explained this to two other outlets. Neither printed it.

"I don't think they were being dishonest," she said. "I think it made the story shorter."

What did not increase

The famous post-2019 rise.

Filed reports went up 38 per cent in our dataset after 2019 and 41 per cent in the public catalogue. That is real, in the sense that more people wrote in.

What went down was everything at the post. On the contractor's own figures, total observing hours fell 54 per cent, and Line 9 now buys fewer instrument-hours than at any time since 1974.

What went up, from zero, is the number of hours in which two instruments are lit at once: 1,095 a local year.

Nothing is arriving more often. Something is being lit twice as often for ninety minutes, twice a day, because a budget was cut and the people left had to brief each other.

What this does not settle

It does not settle a single object.

The roster explains when people looked up. It does not explain what was above them, and this desk would rather say so plainly than let a clean chart do the arguing.

Our negative control did not come out clean either. Run against the primary-return anomaly log PX-1330, the same test finds the same periodicity at roughly a third of the strength, and we cannot tell you why.

Wilhelmina Torreblanca-Ash, who keeps the public catalogue and gave it to us, read all of this and wrote back to say we have proved a fact about a timesheet and presented it as a fact about the sky. Her letter runs below, unedited. She is at least partly right, and the part she is right about is the part we put in the headline.