The geometry doesn't favor it. The worst-case geometry is that everything slumps and congeals in a mass with minimal surface area, which would be a sphere. A perfect sphere is unlikely to be formed. But it you use a realistic term for heat source, a perfect sphere isolated in ambient air, the radiative heat loss is such that you don't approach the melting point of uranium oxide, much less vaporization.
If you don't have optimum geometry (non-sphere, or a congealed irregular mass, or puddle), then the surface area is larger and you get greater heat transfer.
You can take further credit for conductive heat transfer since the material is in contact with other things, and also convective heat transfer, which is always operative except in a vacuum. It could be relatively poor, as in a stagnant atmosphere, but it will not be zero. That further reduces the mass temperature.
Sounds reasonable. I would think you also get (maybe) enough contamination from the rods themselves during any melting, to throttle reaction among the pellets. Otherwise, why go to the trouble to build the the rods that way.
I'll keep my fingers crossed and regularly pray just to stay on the safe-side. ;)
I'd appreciate a ping or freepmail if you come to a different conclusion later, or if something unforeseen arises.
Thanks,