My understanding is that all MGs at least so far use permanent magnet motors. These have the characteristic that above a certain speed, you have to use a thing called "flux weakening". This is an active process, sending currents through the motor coils, that has the effect of cancelling some of the magnet's "strength" (its magnetic flux). This reduces torque, but causes less back-EMF to be produced. This is part of how electric motors perform electric gearing; they automatically trade torque for speed to end up with largely constant power above a certain speed. Without field weakening, the back-EMF would force its way into the main battery, causing the motor to regenerate down to a speed where the back-EMF approximately equalled the battery voltage. Voltage and Electro Motive Force (EMF) are essentially the same thing.
So above this certain speed, and that speed depends on the battery's voltage and hence state of charge, there is no such thing as complete free wheeling. If everything was perfect, the field weakening currents would cause no net power from the battery. Instantaneously, power flows from the motor controller to the motor, then back again into the motor controller. But permanent magnet motors are a bit "coggy"; there is torque ripple from the magnets passing by pieces of iron in the motor's stator. I thought that this certain speed was typically about 60 km/h, but it could be that in a more powerful car like the MG4 Xpower, this speed is more like 60-70 mph. So the vibrations could be a result of the motor's inherent cogginess coupled with the need for field weakening above a certain speed.
As for why ECO mode has less vibration, you have the torque ripple from two motors compared to that one one. It could be that with one motor, this torque ripple is fairly negligible, as it's a fairy high frequency. But with two motors, if they are not perfectly synchronised, the ripples may cause a lower beat frequency, which being lower in frequency may be more noticeable. If that's the case, the fix may be to send CAN messages between the two motor controllers so that they synchronise their field weakening pulses somehow. My understanding is that this puts a constraint on the exact motor speeds, which may cause problems with wind-up as the effective tyre diameters will never be perfectly equal. So they may have to allow the motor speeds to drift a little now and then to prevent wind-up. So it likely isn't a trivial firmware update, and no doubt there is a lot of devil in the detail, and much testing will be needed, varying tyre pressures between runs, perhaps setting up special sensor equipment on test vehicles, etc.
You could argue that this is something that should have been part of designing any AWD EV, I don't know. And of course, I have no direct experience with designing motor controllers, and my theory could all be wrong, or at least largely irrelevant.
.
