Over the days of the Fukushima crisis, attention has switched from reactor building 1 to 3, to 2, back to 3 - and now, to 4.
This is a surprise.
Reactors 4, 5 and 6 were shut down at the time of Friday's earthquake, with some or all of their fuel rods extracted and left in the cooling ponds that each reactor building has under its roof.
Once a reactor is turned off, radioactivity and heat generation in the rods die away quickly; down to 7% of the original power within a second of switch-off, 5% within a minute, 0.5% within a day.
Transferred to the cooling pond, allowing technicians to do routine maintenance on the reactor, the rods are supposed to sit quietly until the time comes for their re-insertion or their journey towards disposal.
The tops of the rods are supposed to be about 5m (16ft) below the water surface.
The water keeps them cool and also blocks radiation.
Over the last few days there have been reports suggesting water levels were low and the water "boiling"; and now the US Nuclear Regulatory Commission (NRC), which has a team of 11 experts advising in Japan, says the pool is completely dry.
This means the fuel rods are exposed to the air. Without water, they will get much hotter, allowing radioactive material to escape.
More remarkably, the Tokyo Electric Power Company (Tepco), which owns the power station, has warned: "The possibility of re-criticality is not zero".
If you are in any doubt as to what this means, it is that in the company's view, it is possible that enough fissile uranium is present in the cooling pond in enough density to form a critical mass - meaning that a nuclear fission chain reaction could start.
The pool lies outside the containment chamber.
So if it happened, it would lead to the enhanced and sustained release of radioactive materials - though not to a nuclear explosion - with nothing to stop the radioactive particles escaping.
The first event in this chain appears to be that the level of water in the pond fell.
Why that should happen is not entirely clear.
Was the fuel so hot that it caused an unanticipated amount of evaporation? Did the earthquake somehow crack the building's structure, allowing water to leak out?
Did the two fires in the building have an impact? Visually, the building housing reactor 4 is the most damaged on the site - suggesting that reports of the two fires being relatively minor were wide of the mark.
Or, did technicians at some point take water from the pond for use in reactor 4's cooling system?
There is nothing to say they did; but during the chaos of the weekend, with power systems and options disappearing before their eyes, it might have seemed like a good idea.
Whatever the reason, the rods became hot enough that steam reacted with the zirconium cladding around the fuel rods, generating hydrogen and causing an explosion.
The same thing happened earlier in reactor buildings 1, 2 and 3 - except that in building 4, the rods were in the fuel pond, not in a reactor.
The government ordered Tepco to put water back in the pool.
But either because of high radiation levels or broken pumps or some other reason, they could not.
Wednesday's plan to drop water in from a helicopter - a technique that is used to fight forest fires - had to be scrapped because of concerns about radiation affecting the pilots. Without the water, gamma-rays travel straight up into the air.
There are reports that the authorities have asked US military personnel to bring in water cannon, which would presumably be fired from the ground, aiming to shoot the water in through the broken roof.
'Rock and a hard place'
The NRC says that in the current dry state, radiation levels from the pond are probably "extremely high", creating a danger to workers at the plant.
Still, in principle this should not raise any possibility of resumed criticality.
According to Laurence Williams, professor of nuclear safety at the University of Central Lancashire, it could depend on how the rods are arranged in the water.
"In some fuel ponds, they dose the water with boric acid at low levels," he told BBC News.
"In some systems they've re-racked the fuel assembly making it possible to put more rods in the pond than it was originally designed for, and then you might put extra sheets of boron in between."
Boron and boric acid mop up neutrons, the particles that sustain the chain reaction.
In this sort of reactor, water is a crucial component of the fission process.
It acts as a moderator - it reduces the speed of the neutrons, meaning they can be captured by uranium nuclei in the fuel rods, inducing them to split.
Without water, the neutrons travel too fast, and are not captured.
Professor Williams raised a scenario that may be unfolding in the cooling pond in building 4. It is just a possibility, because information is scanty; but here it is.
If the fuel rods are dry and hot, there could be damage to the cladding and the release of light radioactive nuclei.
To prevent that, you would want to inject water. But water on its own is a neutron moderator and would enhance the chances, however small, of criticality.
"You're caught between a rock and a hard place," he observed.
Now, Tepco is also talking about putting boric acid into the cooling pond of number 4 building.
How closely the rods were packed, whether any boron sheets were in place and if so whether they were damaged by one of the two fires in the building; these are among the many unanswered questions.
Meanwhile, the most important task remains to get enough water flowing into reactors 1, 2 and 3 to cool the cores.
Following Tuesday's apparent cracking of the suppression chamber in reactor 2 - a likely cause of the radioactivity spike then seen - there were concerns that the same thing might have occurred on Wednesday in reactor 3.
However, the latest reports from Japanese news sources suggest reactor 3's containment system is intact.
If that proves to be the case, the source of Wednesday's radiation spike remains a mystery.
It seems to have been big enough to force technicians to leave the plant.
Clearly, their presence is crucial to stabilising the reactors - and to control the situation in the dry pool in building 4. If they were forced away for long periods, the chances of containing the crisis would fall.
"If water is continuously pumped, they could stabilise the position, because the moment when fuel rods are covered with water, the situation is basically stabilised," observed Jasmina Vujic, professor of nuclear engineering at the University of California, Berkeley.
Tepco sounded a rare optimistic note by saying engineers would soon restore an electrical connection from the national grid - which should allow them to re-start water pumps, provided they have not been damaged by the tsunami or the hydrogen explosions.
While this work continues, so do the questions over the building 4 fuel pond.
Re-criticality seems an extraordinary thing to contemplate; but if it is not a real possibility, why was such an idea floated by the company itself?
The bigger picture, though, is still one of a serious local incident, with minor impacts outside the plant.
US Energy Secretary Steven Chu, however, suggested Fukushima was now more serious than the 1979 Three Mile Island incident in the US - and if contamination does spread outside the immediate area, that will prove to be the case.