The CMRR is more a function of the probes being used than the techniques being shown. The one exception is when using the MATH function on a scope to measure a differential signal, the CMRR is arguably beyond terrible. This is due to inherent phase mismatch that two probes will have from each other, which translates to a mismatch in common mode calculations.
In many of the tests for gate drivers, the exact frequency response of some of the signals is less important than whether there is an extra glitch or timing error present. One important measurement that good CMRR is required is when determining if a Miller turn on is at risk of happening. A bad CMRR could create an exaggerated spike during the Miller current event, causing one to believe a gate voltage is going above a power switch threshold. This extra spike could also be caused by large measurement loops.
