Dielectric resonator Q factor is very high,so why no problems to maintain zero phase?

For DRO oscillators, loop gain is >1, and phase must be ~0 or few degrees. I can't understand, at frequencies say 20GHz, tiny change of 0.01mm make difference in phase. So how it still be working? I can imagine someone moving DR puck and looking onto spectrum analyzer and randomly get some good or better result then glue thing to PCB?

It`s up to you where and how do you glue the DRO. To the metal wall to the PCB directly or via ceramics distancer- pedestal to the PCB or something else.

Problem is why phase maintained right? Isn't it position-related parameter?

Finlay what is your aim, phase or oscillator with wright frequency and performance. Phase and freq depends from everything. Shield material, distance, temperature, position and so on.

20Ghz has 15cm wavelength in vacuum.On a dielectric substrate this is smaller depending on dielectric coefficient.
A tiny movng can create a significant phase shift.
Just a guess..

So, what i want to clear out is: if we have very-high-Q resonator, then it is easy to get s21 maximum out of n*360 phase area, so oscillator will not start. How this problem solved for >20GHz oscillators? Hand-placing, machine-placing of DR puck?

Definitely not in a vacuum. Should be 15mm.

You're right..15mm
On a substratrate , it will be smaller.I meant a tiny movement may change the phase radically..