I've been playing around a lot with ultra-stable clocks in the last couple of weeks, it has been quite a journey. Just to see 9 zeros on a frequency counter running off a GPS disciplined oscillator is quite the achievement and for a long time it looked like it would not work at all.

The next problem to deal with is drift, getting it to work 'for a while' is easy, getting a PLL that is both fast enough to correct and long term stable is quite hard, more so if you want the thing to be physically robust. Even so I am already quite surprised that a few hundred bucks can get you to laboratory standard levels of accuracy and stability. Reproducing that without a GPS to guide you is very hard.