If you really go into the fundamentals, it's possible to make incredibly well-compensated (better than 0.1ppb) MEMS oscillator ICs. I am not sure if they are produced at all outside of research purposes, but I could see devices like that (which have an oscillator which is incredibly closely coupled to a thermal sensing element, gravitational sensing element, feedback loop, etc. etc. all the compensation guts on one IC) eventually being produced for this kind of market. If the frequency stability craze continues a company like Rigol might very well have enough production volume to warrant something custom (but fundamentally cheap) like that.