This stubborn problem threatens the main radiation barrier at the plants: the garage-size steel vessels that cradle tons of radioactive fuel.

In certain emergencies, these vessels would flood with cooling water. If the vessel walls are too brittle, they could shatter and spew their radioactive contents into the environment.

This kind of accident is most likely to occur at pressurized water reactors, the superheated, high-pressure cookers that make up two-thirds of the U.S. commercial nuclear industry.

In the early days, scientists knew that these vessels could grow brittle from years of exposure to the neutrons darting from the core. But they had only vague ideas about how long that would take.

The U.S. Nuclear Regulatory Commission decided to set a 200-degree Fahrenheit benchmark known as “reference temperature” — a calculated measurement that predicts the threshold at which the vessel could break apart. The higher the reference temperature, the more likely the coolant will crack the vessel.

By 1982, 14 nuclear plants had violated the standard. An NRC staff report offered the faint reassurance that no shutdowns would be needed “in the next few years.”

The agency went to work — not figuring out how to fix vessels, but justifying a higher standard.

In 1985, the NRC raised reference temperatures limits to 270 degrees for vessel plates and up to 300 degrees for some welds.

This still wasn’t enough. In 1994, the staff again issued a warning: Vessel welds with high copper content were turning brittle faster than expected. Nine plants might need to close early.

Plants set about rearranging fuel rods to minimize radiation damage to the vessel walls. But reactors kept creeping toward violating the standards.

So regulations again needed to be loosened. In 2007, a top NRC official declared outright that the old safety margins were “overly conservative.” Within two years, the NRC proposed even looser standards for reference temperatures: up to 356 degrees.

These limits were adopted last year, again helping a handful of reactors stay within the rules.

Industry and regulators say these rollbacks are safe. However defended, though, they provide less of a margin for error.

“They do it to protect the acceptable lifetime of a plant and to squeeze more lifetime out of it,” said retired NRC engineer Demetrios Basdekas, who tried to challenge embrittlement standards from within the agency.

The repeated relaxation of the standards might not be enough in the long run.

Per Peterson, who studies reactor safety at the University of California, Berkeley, predicted that the life span of some reactors will still be limited by embrittlement.

And predicting future embrittlement may prove difficult.

When reactors were built beginning in the 1960s, officials placed metal test samples inside the vessels to monitor brittle conditions. Industry experts have warned that the supply of samples is dwindling.