Thank you all for the answers! It gived me idea on topics to search. I read some more about..
I come to some (crazy) conclusions, please connect me if i am wrong:
1)DC bypass capacitor:
Usually connects thin dc-supply lines with ground. At operating frequency of 5...10GHz such lines can be around 100...200 Ohms (1.6mm substrate). As i understand, it would me most effective if bypass capacitor impedance matches between line impedance and gound plane connection point impedance.
What would be ground connection point impedance at 5..10GHz? I think it would be very low, but i am not sure, did not found information about that. So, assume ground impedance is very low, supply line impedance is around 100..200 Ohm. Best matching will occur if we use quarterwave transformer with sqrt(Zline*Zgnd) impedance. As DC bypass capacitor values are pretty high in example above, it is used above SRF, where it start to act as inductor (which equal to 90deg microstrip line as i understand). So capacitor above SRF would work as quarterwave transformer.
From 10GHz examples in my first message most Cbypass capacitors have value around 1000pf=0.001uF, it may be good down to 10MHz and after SRF around 1000MHz will act as inductor (=quarterwave transformer?) .
2)DC block capacitor:
Usually stays on 50 Ohm microstrip lines. Assume we working in X-band with 10GHz sinusoidal signal. If we choose capacitor which SRF is exactly at 10GHz, then capacitor impedance will be extremely low, which makes impedance mismatch to 50 Ohm line. So we need too choose capacitor with higher SRF, so capacitors impedance would be around 50 Ohm at 10GHz.
Here i can't understand why using capacitor above SRF is not recommended in RF. Reactance becomes inductive above SRF, but impedance magnitude is raises very slowly. Also all SRF graphs are given in logarithmic frequency scale, so 0...1MHz we have drop from 100 Ohm to 0.1 Ohm impedance, and from 1MHz to 10GHz(!) we have raise from 0.1 Ohm to 100 Ohm. So why not using capacitor above SRF in DC-block application?