Note that you still get common mode errors, due to the fact that the probe cable has capacitance or impedance to space, and the scope has capacitance across its isolated input circuitry. (No amount of ferrite beads will save you from this.)
Tim
Tim, if you're referring to the image Bravo posted, no that's incorrect.
It's of a truly isolated input DSO where the Reference leads can be placed anywhere without fear of ground loops and furthermore there is no need to use a Maths mode differential setup to do high side measurements.
I know what I'm talking about. And I've performed the measurements, with the pictured instrument.
It is physically, necessarily, impossible to achieve very high CMRR, with a leaded (partially unshielded) probe, and a nonzero length probe body and cable. You can get close, but you can't get perfect.
Many real world, practical situations will not be on the "adequate" side of "perfect", hence this is a real issue. The range of cases where it goes from "okay" to "ugly" is what I laid out above.
(A coaxial probe connector would obviously help a great deal to address this -- but the cable, connector and circuit simply won't be well enough constructed for this to be a truly general solution. Like, it might be okay for MOSFETs, but not GaN FETs.)
A complete solution does exist: if you moved the entire scope input circuit, so that it is enclosed within an ideal Faraday cage, attached to the gate driver itself. The added bulk of the cage will increase the gate drive's loading capacitance on the inverter, but as long as the inverter doesn't mind (for most inverters, it'll be equivalent to some 10s of pF additional load -- no matter at all), the signal received by the scope input circuit will be true, with zero common mode error. This leaves the matter of coupling the received signal down to the rest of the scope, but that's for the scope manufacturer to solve.
Which is actually kind of an interesting idea to think about... you know how Fluke makes a DMM with a detachable face? (And probably lots of others these days, I haven't checked.) Imagine that, but the probe, input coupling, trigger and ADC were detachable! RF or fiber optic would relay the data back to the base unit.
Hmmm.........
Tim