Don't get the wrong idea, PMF and foam will bend, in fact it is quite flexible. It is also quite impressive how well it returns to its at rest state.
Here is a piece of 1-1/2 inch thk blue foam with 5mm Okoume marine ply glued to the tension side.
It's as wide as my windows (the window plug I cut from one of my side doors), so maybe a 24 inch span x 15 inch wide (don't recall exactly at the moment). It's propped up on 2x4s, so 1-1/2 inch off the floor. That's my considerable girth self (about 265 lbs... or about 66 bricks) standing on one foot in the middle. There's about 1 to 1-1/8 inch of deflection.
To this day I still use that piece as a nice flat cutting board to save my work bench. If I had glued Okoume to both sides I doubt it would have deflected 1/8 inch (a guess). If it was just the plywood, I'm certain it would have been all the way to the floor.
I made some test samples from 3/4 thk blue foam, some with 10oz duck and TB2, and some with 1 and 2 plies of 6oz FG cloth and epoxy.
Surprisingly the PMF and 2 plies of FG weighed exactly the same (before filling weave). Also surprisingly, with just one side skinned, I could bend either 10x12 inch sample back on itself... the FG/epoxy remains relatively flexible in such a thin application. (Some people think it will crack and shatter, so I thought maybe it wasn't fully cured to ultimate strength, but after months I can still bend it and it still comes back w/o delaminating. The big reason I decided to go with the FG/epoxy was durability and puncture resistance... it takes a pokey tree branch hit better.)
Yes, driveshafts are tubular to save weight (and cost of material). They don't generally see bending or shear loads (except for in off road situations where they are being driven across rocks
). The math proves that since the outer fibers are under the most stress, they will be the ones to fail first, so they govern the design limits in torque. A solid bar would be much stronger in shear, but that is a much less important design parameter in that application.