Can someone explain why they are building a housing for a mirror rig rather than just making the housing the mirror rig?
I can understand the Atom rig in a housing for a very few types of Hollywood/feature mid range shots but it seems to me that for most actual underwater 3D you will either shoot as wide as possible or shoot macro. For wide you want as wide as possible so might as well build a housing for the largest practical mirror and lenses, and by definition it is going to be pretty big with large volume. Seems like adding the rig in there just adds more size and complexity without much added utility. BTW Pedro and others, are you planning to converge on your full wide angle (i.e. 14-16 or whatever the widest is) shots or just go parallel?
On the other hand, macro shooting doesn't require a big mirror since you are using telephoto lenses and very small IA, so while you could do it in the same housing you are dragging around a huge unit to shoot the smallest subjects which creates problems getting into small spaces, lighting small close subjects, spooking the subects, etc. Seems like it makes more sense to have a separate small housing for macro?
BTW Pawel is spot on regarding the physics of underwater, it isn't a lens issue - it is the fact that you can't really correct for the problems on two separate cameras because the corrections require you to align them with the lenses so if you are aligned on one you aren't on the other. So I agree you should do some aquarium tests to understand the issues before you spend $160K on Atom plus housing.
I haven't done the charts but I don't really doubt the major loss of resolution - it is consistent with the physics. Whether it is "good enough" is a different question, but it may mean there isn't much point in spending huge dollars for no increase in quality. Many of the issues with both dome ports and flat ports are mitigated a bit by the depth of field you get with the much smaller 1/3-2/3" sensors stopped down to f8 or f11 that most of us have used over the years but are exacerbated by the APS-C and APS-H sized sensors in the Red cameras and larger apertures with 3D.
Some of the aberrations may be correctable via software like they do with stills, but I don't how practical it is for 5K motion.