I didn't want to invest in a motor drive for the capacitor until I had completed the experimental stage and I am glad I didn't. The first thing I discovered is that the used capacitor I bought either doesn't make it up to 47pf or something in the loop or calculator is "off." The loop would tune down to 14.150, but not below that before I ran out of shaft rotation. I constructed a coaxial (RG-8U / 26pF/Ft) capacitor to pad the vacuum variable and it would then tune down (max cap) to 13.5MHz on the MFJ-269 analyzer. Here is another note: the real-world dielectric strength of the standard Belden 8214 is NOT up to the voltage potentials of these antennas beyond QRP levels. (Baby, light my fire!)
Did it flame out INSTANTLY or did it take a second? It might be the losses and subsequent heating in the coax stub that get you, not the ultimate voltage standoff of the cable. A foot-long stub of RG-8/U used as a capacitor only has a Q of about 1500 (nearly 0.25 ohms resistance and 380 ohms reactance, abt 29pF equivalent). While that Q is maybe comparable to a cheap sliding-connection air variable, it's a far cry from the Q of the vac. variable.
If I keep the vacuum variable capactor rather than building a new "trombone" cap, I will need to find an 8 KV shunt cap or fabricate one from brass and dielectric.
You could also think about doing a loop-lengthening transplant of longer legs in two sides, make it into a somewhat oblong shape. You'd want to keep symmetry about the feedpoint and capacitor but you could either make it wider or taller depending on where you need to hide it. This might give fewer headaches than trying to fabricate a HV capacitor.
Also, the drive motor input to the vacuum variable capacitor shaft will need a resolution of about 1/2 of 1-degree and sufficient torque (high resolution stepper)
Geared down stepper might be nice.
The 1" (1.125" OD) copper pipe, 45-degree elbows, copper plate for the capacitor mounting plates, and jeweler's silver solder are bloody expensive these days. So, there is considerable effort expended to build and optimize the antenna, and the cost of materials is very high compared to a highly loaded stealth antennas of other configurations.
Yeah, it should be straightforward to do better for efficiency with considerably less cost if you don't have such a low radiation resistance. A loaded dipole that fit in the same bounding box will have considerably higher radiation resistance and not so much attention needs to be paid to losses. The downside is that it's hard to tune.
I'm working on a remotely tunable loaded dipole but... well, "working on it" is kind of a euphemism for "some pretty complete bits are sitting in my shed waiting for me to build the second capacitance hat."
I do like my magloop though. Mine isn't such fat tubing as yours and it's still an eminently usable antenna. I happened to have a 300pF variable and built a four foot octagon: I can get all the way down to 40m and have had several very satisfying QSOs with good reports vs. a much bigger antenna. Not BETTER of course, but roughly aligning with model predictions (good agreement would require tons of measurements) and certainly very usable.
I actually have had more fun using it than some of my big antennas just because it adds an extra flavor to trying to work a DX station on 30 or 40 or whatever. I get a real kick out of using a four foot octagon. A couple years ago, I didn't have time to actually go OUT to field day, so I set up a chair, FT-857, boat battery, and the magloop up in my own backyard and operated that way. I carefully rolled up my nearby 40m half-wave vertical to detune it sufficiently. I didn't want to be a "parasitic cheater" like I was on one of my early test QSOs with the loop
I do think the price of doing it right is quite a bit steeper than it needs to be for the performance and I think it's probably worth thinking about alternatives, but I'm still glad I built it. I'd wanted to try one ever since I first read about them.