Wow, that's probably one great way to spend a little time. I learned most of what I know after I thought I knew it all.
During a fault, amperage goes very high, very fast, that means your 'instantaneous' impedance changes dramatically. Soares illustrates this far better than I ever could here. That is the reason the code does not allow (except in some rare cases, as you quoted) conductors to be separated from each other. In DC it makes no difference, in AC it can raise impedances (with high amperages) to points where it will not follow the intended path back to XO, which is where we want a fault to travel.
The code recognized that adding a ground to an outlet, it is safer to have one, even with high impedance, than to not have one, at least we have established a reference point. remember, that's a contingency, you would not get away with it in a new installation.
Short answer, to repeat Soares, impedance goes high in a fault, always. If you do something to raise impedance such as separating conductors, you make it impossible to have a viable ground path, so never separate conductors.
Magnetic path, in a fault, nearly every metal object within reach of the conductors will have a voltage and amperage imposed upon it, detracting from the fault clearing cabability of the OC device. It's best if we eliminate those as much as is practical even though that is very difficult.
One of the reasons I really don't care for the practice of allowing a metal 90 in the underground portion of a PVC run, during a fault, we waste precious amps in turning this into an electro-magnet....literally. If it is embedded in concrete the problem can be multiplied by that factor as well.
Uh, geeez, did that answer your question at all?