There's no possible way to build an un-trackable satellite that's large enough to make useful Earth observations. The basic limits of physics still apply. There's a difference between rocket science and magic.
I did some calculations, and a 15m object (largest that fits in falcon 9) at 300km (lowest orbit) has an angular diameter of 0.006 degrees. That amounts to taking up about 0.000005% of the sky above the horizon.
I figured the required resolution to cover the sky well enough to see that would be ridiculous, but 20 megapixels spread over the entire sky would then be enough that this satellite could occlude an entire pixel.
Considering how many hobbyists there are, this does indeed seem feasible.
The difference is that they attempt to make themselves look very small in the types of radar used to aim missiles. If the error bars on the location is large enough, you can't fire a missile and get a reliable kill.
The term I keep hearing is the "kill chain", in that stealth is attempting to break one of the parts in the chain.
Satellites also need power. Solar panels are big and have to point at the sun so no way to hide those. RTGs are also an option, but even with those there's going to be significant waste heat which shows up on IR sensors against the cold background.
And any large satellite will show up clearly on a regular optical telescope. Take a series of pictures over time and you can spot anything moving, either through direct imaging or because it occludes some other celestial object.
True, but it would be very difficult to 1) observe this unless you were looking for it directly and 2) determine the object's orbit. Since stars are point light sources, you can't even tell how fast it is moving since the size of the object could mislead your attempt to deduce this by timing the event.
If Zuma made it, and it's not found by the people looking with the right tools, wow. But I doubt it. It's too hard to hide with standard assumptions.
I would love to see the function for magnitude vs optical cross section, no doubt one can calculate that from first principals. It's just observable star density.
> An adversary can (presumably) almost immediately target that location if they are willing to admit they saw it
You'd be late to the party. If it was in orbit around the Earth, and since you don't know its direction or distance from Earth, your impact object would have to do a lot of searching once it got into space. If you had a giant laser to destroy it and could fire instantly upon detection, it will have likely moved quite a bit in the fractions of a second that it took the light to reach it. And, again you don't know the direction it is moving.., you can't compensate.