This is nice causality question.
For the theorist with 99% knowledge of complete theory of electricity and missing few pieces the method of obtaining a reliable high frequencies source will be a mistery. If such theorist will create a Spice software, the models will be perfectly working except models of oscillators.
1) It is self caused
The crystal oscillators are impossible to run without help of thermal noise. Crystals are just linear filters with very narrow band. Amplified signal in MHz range is not the AC, but thermal noise of all components in oscillator. Which is very broadband noise. The exception is quantum based generators: atomic clocks. (but not quantum amplifiers: lasers). Fortunately AC signal from power generator helps to make start of oscillaton somewhat deterministic. (For simplicity consider that power supply is not a switching kind)
2) Use chaos theory coupling
Second question is how is it possible to exploit the existing source of oscillation when creating your own causality root. Or how is it affecting it, and perhaps how to isolate effects or control degree of effect.
This class of problems boils to coupling of oscillators. If they are build in same universe, then they are coupled. There is no symmetry in coupling. AC of 50-60 Hz is affecting local oscillator more than oscillator affects generator of the power station. The effects backwards involve enormously small numbers, but they exist.
For all practical means power generator is not affected by oscillator. For simplicity: It is only one way. The effect causes phases to be coupled with some strength. The strength is rarely considered advantageous, because designers want their devices to be deterministic. However, no matter how hard one will try to exclude the coupling, it will always exist.
Consequently near 100% of crystal oscillators powered on particular continent are participating in one very broad and loosely coupled causality tree rooted in NIST.
When branches of the tree are tangled, the effect is stronger, especially on close frequencies. If observer will measure phase differences and resulting frequency walk between neighbouring oscillators the 2D picture over long time will show more than one cluster of preferred values with abrupt transition between clusters. This trajectory is nicknamed strange attractor.
Problems of indeterministic behaviour of deterministic systems is studied by chaos theory