A common problem, especially with installations where there is one drive feeding multiple motors, is that the cables going from the drive to the motors can exhibit the properties of capacitance between them because of the steep rise time of the PWM pulses. The longer the circuit length the more likely that is to happen, but often times the manuals will only state that as the distance between the drive and motor. On multi-motor applications, that circuit length is multiplied by the number of motors. So if, for example, the maximum stated distance from drive-to-motor is 100m, but you are only 60m, you think you are fine, but in reality your CIRCUIT length is 120m.
So what happens is that the capacitive charging current in that circuit APPEARS to the VFD protection algorithms to be a Residual Current Ground Fault, meaning Ix amount of current is flowing out, Ix-n is returning. The assumption in the algorithm is that n is current flowing to ground, when in reality it is just charging up the capacitor you have created in the conductors.
The cure for that is a load reactor, because by adding inductance to the circuit closer to the VFD, you slow down the steep rise time of the PWM pulses and help reduce the capacitive effect in the cables.
If it was working fine and suddenly began exhibiting this behavior, then look for something that has changed. One thing that exacerbates this phenomenon is when people mess with the Carrier Frequency in the VFD so as to make the motors run "quieter". Increasing the carrier frequency above 10kHz doesn't really make the motor stop whining, but it does move the sound frequency out of the range of human hearing. However at the same time, it will INCREASE the capacitive coupling effect in the circuit conductors.