Okay, so I completed this mission. Here is how it went down.
Getting this vessel into orbit was more like launching a shuttle than a rocket, except the shuttle used none of its own engines. We needed a careful balance of thrust and mass due to the unavoidable asymmetry. Thus, we employed a Mammoth and a Mainsail, side-by-side, with four Kickbacks and a couple of drop tanks. I found that during ascent, the shifting flow of fuel would make control more difficult, unless I progressively reduced the thrust output of the Mammoth to the point that the SAS could compensate for it. Fortunately this was not too difficult as I could generally make the adjustments between burns.
There was an extra fuel tank inside the cargo bay. Since that space would be needed later in the mission for the part we needed to recover, I figured it was a nice place to stick a drop tank we could burn off to make our Munar transfer. Once we were on an impact course for the Mun, the empty tank was dropped, then the engines made a fine adjustment to put us into low Munar orbit, while the tank sailed on to be safely disposed of by high velocity impact with the regolith.
While I had closed the cargo bay hatch after ejecting the drop tank, I was a little worried because by the time I got to Munar orbit the batteries were nearly drained. It turned out that because I had adjusted the placement of the solar panels so they would fit more flush with the surface of the bay, their center of mass was technically inside the bay walls, and thus the panels were inoperable while the bay was closed. Thus, the bay spent most of its time open. In the picture above, the craft has been met at the Mun by a refueling drone because while I had planned for enough delta-v to get to the Mun and back again, I forgot to account for the fact that the place I needed to land had a big inclination adjustment and was getting nervous about fuel levels.
I forgot to get a picture of the lander coming into land, but it has a set of four Twitch engines on each outboard pylon. Not much thrust for this amount of mass, but good enough for the final descent into a belly landing. The landing was successful, a few hundred meters from the target, and it was revealed that Lizke Kerman was there with her Hitchhiker Storage Container. She planted a flag to mark the occasion of her rescue, and the rover undocked from the cargo bay and headed out to meet her.
Fortunately, the rover easily grasped the Hitchhiker, and Lizke hopped into the passenger seat, ready to get out of here. I was a little worried about the rover carrying all that, but it handled it easily. Having that big reaction wheel under the seats (and set to SAS only) helped go a long way toward compensating for any center of mass issues brought on by oddly shaped cargo.
The moment of truth was getting the rover back into the cargo bay and docked in place, and it worked perfectly. Rolled right up the ramp. connected to its docking port, cargo container still snugly grabbed with the AGU. Lizke went into our spacious passenger compartment, grateful to get some fresh snacks and a chance to strip out of her space suit and towel off for the first time in days. Okay, time to get out of here. Doing the pre-flight check, stowed cargo, check, crew strapped in, check, electrical systems, check, engine activation, check, fuel tanks... -uh oh.
"This is Mun Recovery One, Kerbin, we have a problem..."
Turns out that even with my refueling, the craft no longer had enough fuel to make it back to Munar orbit. It could only manage a hop, at best. The earlier miscalculation was worse than we expected. The craft would need to make an on-surface refueling, which is much more complicated than an orbital rendezvous refueling. The engineering team at Kerbal Space Center flew into action, and slapped together a probe on wheels which could land itself, drive up to the lander, and refuel it from below. It was promptly put in a launch vehicle and sent to the Mun.
The refueling drone had two tanks worth of rocket fuel to deliver, making the final stage of its descent on a quad of Puff monopropellant engines. I have no idea how two of the tires got blown up while in flight (I certainly made sure they were intact at launch) but fortunately it was built with redundant tires in mind for just such a situation. The two small ore tanks on top of it are simply ballast, intended to adjust the center of mass both for the sake of the launch vehicle and to make the landing controllable. They can have their content jettisoned once it reaches the surface, but the extra weight helps add friction to the tires.
Unfortunately, it turns out that the engineering team at KSC assembled this refueling drone a little too quickly. It was not able to wedge its refueling port under the lander to meet the refueling port on its bottom. The crew of the lander are now grumbling about shoddy government work and lack of standardization, wondering what they are expected to do with a set of badly needed fuel they cannot actually connect to. Then one of them gets an idea: the rover they brought along is designed to dock with literally anything because they did not know exactly what they would have to recover. The rover itself has no rocket fuel tanks, but it is designed to connect to the lander with an arbitrary object docked to its other end. They spring into action!
They roll the Hitchhiker Storage Container back out of the lander and gently set it down out of the way. Lizke Kerman volunteers to stand downhill beside it and make sure it does not go rolling away. The rover rolls out to the refueling drone and neatly grabs it. The turning radius is pretty wide with this docked setup, but it moves over the Munar surface more easily than the refueling drone alone did.
Getting the drone into the cargo bay while attached to the rover was a challenge. I had to let the AGU freely pivot so it could handle the change in incline of the ramp, but it made it in. It hung out the back, but the rover was able to dock without losing its connection to the drone. From there, it was just a matter of manually transferring the fuel into the lander. Once the drone had fulfilled its mission, it was rolled back out, its brakes engaged, and its type changed to "debris" though I expect it to still remain technically "active" so long as it gets sunlight. The rover connected back up to the Hitchhiker Storage Container and brought it back in again. The whole crew sighed with relief: they can finally get out of here.
We had enough fuel now to aerobreak at Kerbin. Sadly, it was not enough fuel to put us down right next to the KSC, but we were not exactly in a position to be picky. The cupola module very nearly overheated during the first pass, to the point I evacuated the pilot to the crew module behind it in case it blew up, while the autopilot system (which was technically mounted in the docked rover so it could operate autonomously if necessary) kept it going. I threw it into a spin to shed heat, which seemed like just enough to keep the cupola from exploding. Future passes were done rear-end first, which was able to better handle the atmospheric compression. It had just enough chutes to bring the whole thing just under 10 m/s when fully deployed.
Touchdown! We have returned to Kerbin and are ready for recovery.
Recovery contract complete. That was actually pretty fun!
Fearless Son's post in Recovering parts from the Munar surface was marked as the answer