isolators were WIDELY used back in the day.
but they have a fixed cost (at least $40 USD in quantity, and often much more), their size is HUGE compared to modern system-on-a-chip applications, and they often do not have a very broad bandwidth. Hence they fell out of favor.
places where you still might need one? a transmitter with an outdoor antenna that can ice over. the heavy ice will detune the antenna, which will reflect the transmit power back and blow up the amplifier output if allowed. an isolator with an appropriately sized load is one solution, as is a reflected power detector that will simply turn off the transmitter.
So for that application, you need to handle high powers with a big standing wave.
Other applications, lets say you have a system that needs tight bandpass filtering. say it is a downconverting receiver, where you are trying to remove LO and image frequency leakage. You typically can build a tight filter with good rejection into a 50 ohm system. but the MMIC mixer you downconvert with might have a poor vswr, and an amplifier on the other side of the filter might also have a relatively poor return loss (perhaps 10 dB). that can detune your filter, causing all sorts of weird digital modulation distortions due to amplitude or phase non-linearity. Available size and cost preclude you from using a ferrite isolator.
so if you can make a physically small isolator equivalent, that has flat group delay, good return loss, low cost, low amplitude and phase ripple, i could see putting one between every mixer and an adjacent bandpass filter. Be aware, you need bigger bandwidth than you might think. lets say you are receiving a 9 to 10 GHz signal, and using 8 ghz lo.
to reject the image frequency, your isolator would need good return loss from perhaps 5 to 11 GHz, for the bandpass filter to work as promised.
if your method of non-reciprocity includes nonlinear effects, such as am-to-pm conversion, am clipping, harmonic or third order product generation, it might make it less useful for some systems.