Hi, that was an interesting method. Of course, they had little choice but to try something, as the main communications array was down, and the back up array ran so slowly that they were recording data on a 4 track tape for "time shift data telemetry". The LEDs under discussion were the ones for the servo drives in the tape deck, which had failed.
When we use LEDs, you sort of see this effect already. While not as prominent of an effect as it used to be, we used to (5 years ago) burn-in the LEDs with a 24 hour run at full power. This could knock down the Vf as much as 0.5 volts. Modern LEDs are much more consistent, so they already come with the reduced Vf.
The crystal structure that was "repaired" was a GaAs / InGaAs type, which anneals at much lower temperatures than GaN, and those were also a much simpler device structure than what is run today.
I suppose we could run a few LEDs through Fukushima and test it.