I was hoping to start fab by now, but I ended up having higher priorities this month.
Anyways, I did some CAD to help firm up my planning. Might as well share the aero eye-candy.
The lid ended up with so many compound curves that it would be very hard for me to learn how to make it - and make it look good - in aluminum. But the good news is the lid shouldn't need to take as much abuse as the rest of the box, so fiberglass should be great. I'm going to make a mold for the fiberglass with plywood and foam.
The rest of the box will be aluminum sheet. The the only compound curves will be contained within 6 corners. All 6 have a 4" radius, so I'm thinking that the stump method would work well. I see one or two spots, other than the floor, that probably need some framework of one sort or another. I figure I'll want a buck, at least for a few cross-sections, as well.
The 6 shrunk corners shouldn't be too ugly, even if I do them poorly, since 4 on top are covered by the lid, and the 2 on bottom might stay out of the way. This plan should pretty well cover wife certification, give or take a little bondo over the rivet heads, and it keeps my tool outlay pretty minimal as well. Not that I would really be sad if I ended up with some sweet metal shaping tools, but I'll try the stump and probably some dollies and a slapper and then go from there.
Planning on a metal square tube frame floor with pink foam in the gaps and 1/4" plywood over the top. This gets bolted to the inside of the rest of the box.
I've depicted circular tube cross members - I was hoping to find off-the-shelf lugs and do silver brazing, but I can't find generic lugs anywhere. If I drill through square tubes then the load on the joints is probably fine with just epoxy putty to keep them from rattling. But I might change my mind and get a welder and do everything in 1.25" square tubing. Man, that would be fun.
There is also the caster shown; I still haven't decided for sure, but I think this is the way to go. This design means in a ground-strike the caster and hitch receiver will have to bear the weight of some of the vehicle. The caster here is rated at 300lbs, so it may or may not break. My experience is that the car will not strike at this departure angle if it's unloaded. So if I do this right, the weight on the caster should be no more than the load in the car, not the car itself. I've also set the bottom slope at 15.5° to keep the departure angle sane, which is faster than ideal but will still provide some aero benefit.
Another approach is to use a caster, hinge and chain so that the box can just bend out of the way instead of trying to support the vehicle. This would also enable lowering the bottom slope closer to the aerodynamic ideal. But space is tight between the trunk and the box, and it adds complexity, and I'm debating whether it's worth it, since I'd have to redesign everything. (Better now than later though.)
EDIT: I forgot to mention; this box has about 18 cubic feet for interior volume, says my CAD software. That's gonna be great! The hitch rack / bag I've got now is 11.5 cubic feet. The Civic only has 11.9 cubic feet in the trunk.