Inside an MRI magnet, current flows through a closed superconducting loop, like water circling in a bucket after you give it a spin, it just keeps going on its own. To change that current, you heat the loop with a small heater: it stops being superconducting, becomes an ordinary resistor, and then an external source can 'top up' or 'draw off' current. The usual assumption is that once you turn off the heater, the loop instantly returns to its superconducting state, like water freezing solid in a split second. But in a real magnet, things are different.
The scientists measured the loop's resistance in two modes: with the heater on, and after turning it off when the loop heats itself. It turned out that in the second mode, the resistance is 12 times lower than in the first, at the same voltage. But the key point is that both modes are described by a single simple formula: resistance depends on the heating power and the current through the loop. It's as if the kettle's temperature depended only on how much heat the stove supplies and how much water flows through it.
This behavior affects the accuracy of the current in the magnet: the longer you keep the heater off, the more precise the current. In 18 maintenance sessions of MRI scanners, the scientists observed that after turning off the heater, the current in the loop barely changed — the residual current was only 8 millionths of the main current. That's like having just a few drops left in the bucket after the water stops swirling.
The switch's resistance changes by a factor of 690 when transitioning from the cold superconducting state to the hot resistive state, as if a garden hose suddenly became as thin as a human hair.