Yes. This has been done in the lab to improve the phase noise of reference oscillators, though it gets expensive in parallel hardware if iterated.
Start off with two independent VCXOs of the same frequency. Use a mixer with a DC-coupled output to combine their outputs. The sum frequency is the doubled output. The difference frequency, their phase difference, is used to phase lock them together. The output of the loop goes to increase the frequency of one, and decrease the frequency of the other. The combination can still be voltage controlled as a single entity, by using a tuning signal which increases the frequency of both.
When frequency doubling a 50 MHz oscillator to 100 MHz in the conventional way with a doubler or PLL+divider, the phase noise increases by 6 dB. When adding two 50 MHz oscillators together like this, the phase noise adds as power, so only increases by 3dB. That's only a 3dB improvement, for the expense of two oscillators and their control circuitry, but if you are grubbing around at the margins of what is possible, it's 3dB you won't get any other way.
You could combine four oscillators for 6dB improvement at 200 MHz, and 8 for 9dB at 400 MHz, though you now need to duplicate rafts of hardware for each further 3dB improvement.
It is possible to get the same effect without phase locking the two VCXOs together, if they are both phase locked to a common reference. Adjust the relative phases of the two sources at the mixer, perhaps by changing cable lengths, to maximise the sum output. This also brings the DC output to zero, which is perhaps a more convenient way to monitor their relative phase.
Of course if you are in a lab and can borrow a couple of signal generators, you can play with this effect. Make sure both are referenced to the same 10 MHz external reference, set them to the same frequency, mix them, and adjust their relative phase. This is rarely the best way to get a 2x frequency signal, or a 3dB phase noise better signal, but it's instructive to do it once.