Chiming in here just because I happen to have taken a handful of online-proctored certification exams through Pearson OnVUE recently.

The gist of it is that I think someone willing to put a lot of work in could probably cheat using the strategies you suggest, but it would be a pain. During your checkin you have to send a selfie, photos of the front and back of your photo ID, and four photos of the space you've prepared. You can't have any writing tools, written materials, or anything else that looks like a computer or screen in the area, and the machine you're on has to be single-screen. If the pre-test greeter or the proctor aren't satisfied with what they see they can ask you (via text and voice chat) to show them the room in real time via your webcam and may ask you to make changes or move things around to provide evidence that something in the room is not hiding mechanisms used to cheat. From that point on, your webcam and mic are on and live streaming to the proctor for the duration of the test; they don't say anything about assistive technologies on their end but I assume they are using eye tracking to look for instances of eyes wandering offscreen for a protracted period of time. The test environment software effectively "takes over your PC" during the test and I would imagine is pretty effective at detecting alternate/multiple display outputs etc.

There are probably scalability issues, but privacy is not an issue from the perspective of the proctor - you are effectively surrendering it by agreeing to take the test online, you could have gone to a test center instead.

I'm curious what the live video stream captures. Is it you, your desk, and several feet around you (i.e., enough to confirm that there's nothing else near you)? It seems like they need enough to do ongoing, real-time confirmation that your workspace keeps matching the still images you sent in.

FWIW the 'privacy' concern is often voiced by students who find the test proctoring intrusive. On the one hand I agree it's intrusive but on the other hand it seems reasonable for the short-ish amount of time (several hours) that they'll be taking the test.

I'm guessing that some fraction of the students complaining about privacy genuinely object to the privacy issues, some folks may object to having to pay a fee for the oversight, and some fraction are objecting in an attempt to get the oversight removed so they can cheat. I'm sure there's overlap, and other reasons that I haven't thought of.

The scalability problem(s) comes from the need to have a human watching each student take the test. The more people each watcher needs to watch the less effective they'll be, but the lower the watcher-student ratio is the more expensive it will be. Especially since a good fraction of the students won't cheat (so you'll be paying people to watch students not cheat for several hours).