That's the first time I've heard of an FPGA failure shorting then after time recovering (like a thermal fuse). Every failure I've ever seen once shorted, stayed shorted to ground.
Have you tried using JTAG to check to see if the FPGA responds?
Are you programming the FPGA at all?
One possible source of this issue, if you have a bitfile loaded in the flash device, some versions of quartus defaulted to all unused IO as output driving ground. this could potentially explain the short then recover. But only if you have some IO's connected to the 1.2 rail. (Possible, if your board pinout is compatible with all the package compatible devices. Usually the largest device has more VCC_INT than the smaller devices)
It's a long shot, but work looking into.
Pete