well, there is a transit time. If you want to look at the phase of a target, you need the phase noise low enough so that when the return signal hits the receiver, the LO has not drifted so far in frequency (or phase) to make a determination of the return phase and amplitude. If your target is only 2 feet away, then because the transit time is small, you can stand a much noisier oscillator (ie. the coherence length can be very short, o another way of thinking about it is the frequency can drift a LOT of MHz/second if the transit time is only 5 nanoseconds).
If your target is 20 miles away, then you had better have a very stable frequency source (i.e. low phase noise).
In addition, IF you target is small, and has a slow velocity, it will tend to be hidden in the returns from the stationary clutter. So you will want, in that case, to have very good close-to-the-carrier phase noise to improve the subclutter visability.
but in general, you have to simulate all that stuff and see what an acceptable phase noise specification can be to achieve your detection goals.