> Perhaps the MIT drone could use a Stirling engine with a Pu238 RTG heat source and air as the working fluid. If so, it could stay up forever.

Radioisotope thermoelectric generators only output something like 250W up to a max of 600W. Not nearly enough for sustained flight of something this size. Also, if you think the licensing and permitting process is hard for a regular drone, take a look at the amount of paperwork and security related to all of the RTG powered spacecraft/probes launched in the last 30 years. Pu238 is, to put it mildly, a somewhat controlled substance.