Splitting the GND plane is a terrible idea for nearly any circuit, but there are exceptions. Let me come up with one example I think that is an exception, a motor controller.
In a motor control application there are huge currents with high slew rates and analog and digital signals are involved. From my experience it makes sense to split the GND return currents from the motor controller and connect them somewhere to the GND plane where the analog and digital parts are embedded. The most sensitive part of such a motor controller is the current measurement, where the noise needs to be suppressed as good as possible. What I have done in many inverter designs is something like this (here the shunts are in the motor phases):
As long as no signal goes over a split in the GND plane (and also in the power plane), there should be no problem. I never could really compare a design with and without a GND split for exactly the same setup, that would be really interesting.
Theoretically, the HF currents will not flow into the ADC section also when there is a solid GND plane (including the motor control return currents), but the LF currents might. I'm unsure about the influence of the LF currents.
So what is your opinion about the GND split in a motor control application?