Hi BW - Warning technical stuff ahead .....Timber is a bit different to metals and plywood is different again because it is a laminate. The main delta is that it has poor shear stiffness as a material vs metal (isotropic material). If you do FE and tell the FE its an isotropic material you are way over estimating its shear stiffness.. So the FE deflection will be less then the real deflection unless you dial back some of the numbers or do an orthotropic analysis.
The outer geometry (the box) is responsible for its bending and torsional stiffness. So a box is great except where the loads come in and out of the structure. In a cnc gantry this is a moving spot (the saddle) so using internal webs means the beam will be different local stiffness at the webs and where there are no webs. If its an open box the corners are what provides the resistance to lozenging. If the walls are thin then the corner stiffness is low and the section lozenges from either torsion or direct shear. This is a dilemma as some of these effects are neglected in std linear FE (1st order elements) and definitely neglected in manual calcs using beam equations unless you understand second order eqns and that's really complex past my pay range. I use linear and non linear FE techniques that can account for the shear deflection (second order elements)... I've done a lot of timber boat structures to survey and they are picky.
You mention strength but strength is not how to think about it. The stress in these structures is really low and its not a strength issue its a stiffness issue. These things will never break but they do deflect and every part deflects a little bit adding up to potentially a lot of global deflection.
So a filled box or holy box solves the local shear stiffness problem. Gantry beams are not "long" beams or what's called a flexure beam. if you look at the flexure beam equations they are applicable if the shear deflection can be neglected. A beam is considered short if its length is about <10x its major dimension. So a 100x100mm gantry only needs to be 1000mm long and the flexural equation is incorrect as the gantry is shear dominate. Most gantries are shear dominant with some flexure so the local shear stiffness is really important if you want to design at that level. This is the main reason construction extrusions when filled with epoxy granite do better as this type of extrusion really sucks in shear stiffness. This is called shear flow as well.
Getting back to timber. Timber is in fact closer to a foam then a solid. Most timbers are about 50% air and they are highly oriented. That's why they are stiff along the grain and not stiff across the grain. Ply and LVL and mass timber fix this to a degree. If you use FE its worthwhile setting up a deflection test and model that and tune the FE to the test. Then at least you will have the bending correct.
if as you say the box is predominantly torsion and its material has poor shear stiffness aka timber then std equations will let you down in predicting the stiffness. This is because the section will lozenge, lose its shape and be inefficient in resisting the torsion if the box is too thin. If you look at trees they are very thick compared to their diameter this indicates something. Trunks are put under very large torsion and bending and they optimise to this. Being alive they don't like breaking.
So in summary a "box" is not always the solution for a gantry or a "beam". Its convenient but does have some pitfalls if its too thin. Thin is relative to the material and to the local loading. Happy to do some actual FE to explain if needed. Its hard to generalize on some of this, need actual geometry and conditions to figure out solutions. The best approach I have so far to take advantage of plywood is to laminate aluminium sheet on the outside. This leverages AL stiffness and the timber fills in the middle for some light shear transfer/ shear stiffness... Peter