Coherency has to be over a period of time. Instantaneously any two waves (or signals) at the same freq will have at at one time or point a phase difference. If their freqs stay exactly what they had at one time, they will stay coherent with each other, i.e. The phase differences won't change over the same period of time for the two. Over a period of time the oscillator or clock that determines the freq of a wave, will drift and have random fluctuations. If the drift of two such oscillators or clocks, is zero and their instantaneous fluctuations also zero (impossible unless at zero Kelvin) they stay coherent. Small drifts and small amount of random fluctuations will make them coherent to some extent. So drifts increase over a longer time range, and random fluctuations also affects it - the probability density function of the so called phase noise which causes those random fluctuations (in freq and in phase) is highest at the oscillator center freq, and goes down as you go away, so one must calculate.
Usually coherence can be calculated/analyzed by knowing (somebody measured it and specified it, or you go measure it) the stability of the transmitter (usually set by the oscillator), and is denotes
D as eg, 10^(-6) over 100 msec. Means that over 100 msec the difference in freq (and from that you can compute phase differences) is 1 part in 10^6, eg 1 KHz at a freq of 1 GHz. Clearly ok for pulsed Non-coherent detection, but not so good for non-coherent detection (eg, in radar, one would use a signal at that freq and look for Doppler changes, and 1 KHz would be a Doppler you would look for, so this coherency (or stability) would not be good enough in a Doppler radar).
You can estimate the coherency of two signals using standard signal processing techniques. First you must digitize, with a x100 (or dependent on the sigma you can tolerate) time better stability in your oscillator than whatever you want to estimate. Then you do a cross coherency estimate, it is very similar to a cross correlation, both of which done with correlators or Fast Fourier Transforms.