I don't see how this can matter any in a typical residence as the neutral and ground are the same size and as often as not terminate at the panelboard on the same bar.
Correct. They generally are of the same AWG. The issue is that a surge impulse can create a large potential difference along any conductor -- we just don't want that potential to be different on a
grounding conductor.
In many electronic circuits, ground is used a circuit reference among components. Now consider two pieces of sensitive equipment, each powered and grounded by a receptacle on the same or even a different branch circuits. Here's the essence of the problem: Consider the case where two pieces of equipment have their grounded chassis interconnected with coaxial cable, video cable, or audio cabling. In most cases, the outer shielding is bonded to the equipment chassis at both ends of the cable. The shield of the cabling now creates a second surge pathway. Why? Because the 3-mode SPDs on the branch circuits are putting it there.
Now consider an inductively coupled surge on a line conductor that is diverted to the grounding conductor by the 3-mode SPD. Depending on the induction point, a huge impulse of current splits, and flows along the two stated pathways. The conductive pathways are not perfect conductors and as such, voltage is dropped along the grounded conductors, that includes the premise ground conductor and the grounded equipment cabling. The more I rises, the more E is dropped. At the moment of the surge the potential between equipment cases is then unequal, sometimes by a significant amount. That potenial difference creates real-world problems with electronic equipment.
Contrast this with a diversion of the surge only (not a safety fault) to the grounded neutral conductor. We've now diverted the surge to neutral where it is dissipated back at the common ground/neutral bonding point.
For folks who are concerned that the pathway back to the ground/neutral bond can present too much resistance, it's a fact thay the popular 3-mode SPDs already do this. Schematically, you will see an MOV from Line to Neutral, in addition to the remaining two modes (i.e., Line to Ground, and Neutral to Ground). The best solution to deal with this involves storing the surge energy, then bleeding it on to the neutral conductor.
Summary: the best premise surge protection solution involves: (1) 3-mode protection only at the service entrance; and (2) "store and release" 1-mode protection on sub-panels and select branch circuits.
The expensive hospital-grade (HG) Leviton and Hubbell 3-mode suppressors being recommended and sold by many A/V low-voltage installers? Throw them out. They have no use on a branch circuit where they're doing more harm than good. The HG grade is fine, just not with "three modes of protection." Let's keep "three modes" only at the service entrance.
A diagram? This is what I'm talking about:
http://www.zerosurge.com/howsuppressionworks.cfm
http://www.brickwall.com/how-it-works.html
http://www.surgex.com/library.html
-Paul