Most often, this final test is done online, without any supervision. For some classes (like math classes), there is an online proctor required (ProctorU, costs around 15 bucks an exam). You might have your own proctor somehow, I didn't try that.
Grading is another area that I am ambiguous about. Peer graded exercises are kind of hit or miss. Sometimes, the peer marks something as incorrect because they are incorrect. Sometimes, they mark everything as correct because, I don't know, they want to be nice or something. Sometimes, there are mistakes in the grading key that peers should use, and that's the most infuriating to me, because the teachers' attitude towards mistake in grading keys vary significantly (and they mostly correct your grade but not the grading key, so future students will hit the same thing).
Overall, the tests for CS courses were, so far, somewhat unimpressive... I've seen questions like "who created C", or "what is the output of this python code" type of questions. The tests are mostly multiple-choice. I'd prefer if the class could elect not to do tests, and do a large project instead. Maybe for more advanced CS courses, that's how they do it, but so far, I've not seen anything like that (and I started 2 years ago, though I took some breaks). It's mostly the exercises and assignments that students are graded on throughout the course where the bulk of learning happens.
For the math classes, they were much tougher, where even a multiple choice answer is not making it significantly easier. For math classes, the tests are no joke. The strange thing was that the math classes teach quite advanced concepts that require you to calculate, for example, e^x or a sum of squares, or straight out use R, and then, in the final tests, you're allowed only a four-function basic calculator. That's amazingly frustrating to me, and I ended up guessing a lot of the questions (as in, e to the power of 4 would be probably less than answer A, but more than answer C and D, so B it is).
