We have similar problems for a similar project. What you should really worry about is condensation. That means water, that may not be much harmful if you think of DC conductivity or corrosion, but a droplet between two near pins of a high-speed differential pair can become a capacitor and spoil all the signals. So, an enclosed chamber purged with N2, dry air or CO2 may be necessary. Or else, some heating keeping the temperature above the dew point.
About temperature itself, the risk is just if the FPGA can start or not. Once it boots and runs, it will probably keep warm enough to operate. You need to carry out some thermal analysis with thermal conductivities, convection coefficients, etc. They are not easy, highly empirical. If you are not short on energy, I would suggest a controlled heating system (independent of the FPGA, software-less if possible (yes, once upon a time there was smart electronics with just dumb transistors, ask the gray-haired old grumpy ones...). Keep temperature at, say, 15 - 20C if your enclosure has normal (humid) air, or at the minimum temperature for the FPGA to start if you use a dry gas.
If you go for the airtight dry gas chamber, be aware of the pressure difference with the exterior. It's actually simpler an open enclosure but you must keep a higher temperature to prevent condensation, what means energy, so that's quite a key to choose the solution.
Hope that helps!
