The theory of the point of use protector is that it shunts transients within the protector so you need a lot of joules in the rating. Basically you are just heating up an MOV or two.
And everything else. The MOV has a low resistance when the voltage is high, so the dissipated energy is low. Unless a destructive fault current flows (like a direct strike with no alternate path to ground), the MOV is going to hold the voltage drop at the clamp voltage where it goes back down in resistance. The conductors will dissipate a lot of it, too, because it is current flowing. A fuse or breaker in the circuit to the POU protector could blow/trip, as well.
The ground connection is just there to provide a path for that delta. You still should be just burning off the transients that were generated within the building there though. The bulk of the incoming transients should have been stopped at the Dmark or the service disconnect panel where you have a very short grounding path. I would drive a rod there, even if I had a Ufer, just for a local ground reference (bonded to the other electrodes).
I agree.
You can make some fairly good protectors yourself if you carefully match the MOV to the expected line voltage.
Hosfelt will sell you them in all sorts of different voltages and joule ratings.
Yup! But you won't have the advantage of it being a listed protective device and the insurance that comes with some protectors. Your own insurance might not pay out in the event of a surge too large to be protected, claiming that it wasn't so large and your unlisted protector was faulty.
I think I may need to do just that, though, because I have yet found a POU protector for a L-L 240 volt circuit (NEMA 6-15/20P and a few 6-15/20Rs although I'll take C13/C19 as an alternative). 330V MOVs on L1-G and L2-G, and 660V MOVs on L1-L2. I do have some power strip boxes with the traditional looking receptacle in the box in 5-15R form. So I just need to find 6-15R in that form, and rewire the guts for 240V and L+L+G circuit type.