I guess you can aways assume it's spinning around COG. For modelling purposes, if it seems to be rotating around another point, you can consider it both rotating and translating.
Having a front engined car with a forward CoG allowed driving "by the seat of the pants" as the driver, being behind the CoG, felt lateral movement from rotation much more, and so could control it from the feedback received.
In more recent designs, in which the driver is likely to be just ahead of CoG, that feedback is much reduced, reducing the available time in which to react.
I think that while low PMI is desirable, its effect on handling can be altered greatly by the position of CoG. The type of car being considered is important.
A lightweight car is generally better, but how would a lightweight car with its engine behind the rear wheels compare with an equivalently equiped but heavier car (within reason) with a mid-mounted engine ?
Starting with a clean sheet of paper is one thing, but how would you select an existing car, to be prepared for racing, from a variety of layouts that were close in standard performance terms?
If major changes in layout and weight were not allowed, which layout would offer the greatest scope for improvement ?