Every carrier is addressed. You dial the station number and the system routes it. Eighteen stations exist. Something has been sending from nineteen since 2013.
I look after the pod system at a large hospital. Air tube, pneumatic tube, the Lamson, depending how old you are. It is the thing that fires a plastic carrier full of blood samples from a ward to the lab in forty seconds instead of a porter walking it in eleven minutes.
People are amazed it still exists. It exists because nothing else is as good. A hospital that loses its tube system discovers within a fortnight how much of its day was forty seconds long.
Here is the shape of it, because you need the shape.
There are stations. A station is the terminal you load a carrier into, and it has a number, and the number is on a label on the front of it and in the controller and in the routing table. You put the carrier in, you key the destination station, and the blowers and the diverters do the rest.
There are eighteen stations in this hospital. Wards, theatres, the labs, blood bank, pharmacy, ED, ITU.
I know there are eighteen because I have physically stood at every one of them, and because I have the as-built drawings from the 1974 install and the two extensions, and because I have crawled a good proportion of the pipe.
The controller logs everything. Every transaction: time, origin station, destination station, carrier ID, transit time, and any fault.
Since March 2013 the log has had transactions with an origin of 19.
Not many. Somewhere between eleven and forty a year, and I have counted them all.
There is no station 19. There has never been a station 19. The numbering goes 1 to 18 with no gaps and the drawings agree with the controller's own routing table and the routing table has eighteen entries in it.
The first one I saw, I assumed was a corruption. A garbled packet, a bad byte, a station reporting its ID wrong.
So I did what you do, which is look at the carrier.
Because that is the thing about this: it is not a logging artefact. A physical plastic carrier, 110mm, the standard type, arrives at a destination station with a manifest that says it came from 19.
I have handled forty-one of them.
What is in them.
Nothing, at first. The first nine or ten were empty, and an empty carrier moving through a system is not unusual: they get returned to balance the stations out, and there is an automatic empty-return routine that runs at night.
The eleventh had a request form in it.
A paper request form, the sort a ward sends with a sample, and it was filled in properly, in biro, in a doctor's hand. Patient name, hospital number, date of birth, ward, tests requested, signature and bleep.
The ward on it was a ward. It exists. The tests were ordinary: full blood count, urea and electrolytes.
The hospital number was not a valid hospital number for this hospital. It had the right number of digits and the wrong check character, which is the sort of thing that happens when somebody writes one out from memory.
The date at the top was the date it arrived. That day.
I took it to the lab, because what else do you do, and the biomedical scientist on that afternoon looked at it and said, where is the sample.
There was no sample. Just the form.
She said, well, there is nothing I can do with that, and put it in the confidential waste, and I stood there and let her, and I have thought about that a lot since.
Since then I have kept them. Twenty-nine forms across eleven years, and I have them at home in a box file, which I am aware is a data protection incident on paper and which I will deal with when I retire.
They are all like the first. Correct in form, ordinary in content, and wrong in exactly one way each time.
A hospital number with a bad check character. A ward that closed the year before. A consultant's name that is spelled right but who left in the nineties. A date of birth that would make the patient a hundred and nine.
The handwriting is not the same across them. That was the first thing I checked. There are at least six different hands.
Now the part I need to be careful about, because this is the bit that made me start writing it down.
In 2019 we did a full survey ahead of a new build, which meant tracing every pipe with a locator and comparing it to the drawings, because you cannot put piles through a live tube system.
There is more pipe than the drawings show.
Specifically, off the main riser in the old block, at second floor level, there is a branch that goes east about twenty metres and stops, and it is not on any drawing, and it is capped at our end with a blank flange that has 1974 paint on it.
Twenty metres east of the riser at second floor level, in 1974, was the old outpatients block. It came down in 1998 and the footprint is a car park.
Our end is capped. I have seen it, I have had my hand on the flange, and it has not been off: the paint is unbroken across the joint.
So there is twenty metres of dead pipe with nothing at the far end and no way in at ours.
I put a camera down it in 2019 after the survey, through an access port, and it goes twenty metres and it ends, and there is nothing in it.
I have not connected those two things in any way I can defend. A capped branch is not a station and a station is a terminal with a keypad and a blower.
But the transactions with an origin of 19 route in through the main riser in the old block, every one of them, and that is the only place in the hospital where they could enter from the east.
I have not reported it, and here is my honest reason.
The system works. It moves nine thousand carriers a week and it is the reason a crossmatch takes forty seconds instead of eleven minutes, and there is a business case somewhere for replacing it with porters and an app that gets dusted off every time anybody says the word problem.
If I put in writing that the controller logs transactions from a station that does not exist, I do not get an investigation. I get a line in a risk register, and the line sits there until somebody uses it.
So I keep the forms and I keep the log exports and I do my PPMs, and twice a year I stand in front of the blank flange in the old block for a bit longer than I need to.
The last one came in February. It had a form in it and the form had a hospital number on it with a bad check character, and a ward, and a date of birth, and in the box marked tests requested somebody had written, in a hand I have not seen before:
Still here.