While doing some research on making an ultra low distortion analyzer I stopped to look at the Fluke 510A that has been siting on my shelf for years. Lots more on the 510a here: http://assets.fluke.com/manuals/510A_AD_imeng0000.pdf
In short a low distortion (.005% or better) 10V +/- .002% reference oscillator. There are very good numbers and by 1970 standards exceptional.
I was looking at it because I was researching AGC circuit for oscillators. This is an interesting implementation with its standards lab precision accuracy and stability. However its not simple and the critical parts are expensive (an ultra stable voltage reference, ultra stable resistors etc.)
In any case I was wondering if the distortion can be reduced still further. At the time it was built .002% was about as good as possible in practice and better would be hard to verify.
In short, after a number of experiments I got the distortion down to .0003% THD (30 KHz...