A few quick updates on this:
We probably need to abandon the spherical CM idea. When I stated that it subjected the occupant to 8-9 g during reentry, which was uncomfortable but entirely survivable, I was not taking into account the increased reentry velocity for lunar return missions. This is about 3000 m/s higher than reentry velocities for LEO, and this increases the acceleration. A pure sphere will subject its occupant to about 20 g during reentry from the moon, which is well past human endurance. A large ballute could lower this to an acceptable limit but (i) see below commentary on ballutes and (ii) if the ballute were to fail in some way it would be a fatal problem with no hope for recovery, which seems like something to avoid. That said, parachute failure is a pretty serious problem too, but (i) parachutes have a pretty solid track record, whereas ballutes are super new, and (ii) I think multiple redundant parachute systems are likely to be much more achievable than multiple redundant ballute systems, from a mass/volume/cost perspective.
As for ballutes: I have read a few papers on these now and am not quite so enthusiastic about them – for CSTART. They are a fantastic idea in general and I think they show a lot of promise. But they are by no means cheap or simple yet. Most papers stress that for ballutes to become practical as routine and reliable devices for atmospheric reentry, some advances in materials technology will be needed. Ballutes require fabrics, plastics or metallic films (or layered composites of more than one of these) which are simultaneously very low density but still able to endure thermal loads which exceed the tolerance of most off-the-shelf materials. The first working ballutes for something like lunar return reentry are probably going to be made of multiple different materials at different parts of the ballute, and all of them are likely to be composed of several layers of unobtanium). This doesn't make ballutes completely impractical for CSTART, but it seems clear that parachutes are going to be a heck of a lot easier overall.
With regard to replacing the spherical CM, I have started to think in terms of a "headlight" shape, as seen on the Soyuz reentry module (see here for a nice diagram). This has an L/D ratio of about 0.3, which is enough to significantly lower reentry g forces, still has a fairly good volume to surface area ratio (it was designed to maximise this ratio, subject to providing adequate lift for lunar reentry), and I think it might actually be easier to work with, e.g. attaching the landing legs/engine/tank assembly to it's mostly flat bottom will be a lot easier than attaching the same thing to the base of a sphere. We do lose the ability to do independent translation and rotation using a single set of thrusters, which is quite sad, but this might just be something we have to accept.
I have a few more thoughts brewing on how to configure all these bits of hardware, but I'll hold off on posting them until I have given them a little more thought.