When I was an honours / postgrad mathematics student, I used to run some tutorials for first-year students. I always checked line by line and made sure they didn't try anything like this. And yes, I checked carefully across page boundaries too.

If there seemed to be an unsupported jump in the logic from one line to the next, they'd lose one or more marks (depending on how big the gap was). Some students were honest about it, some tried to hide it. I always appreciated the honest ones. And for every error, I tried to explain what was wrong, why it was wrong and what they should have done instead (if that wasn't obvious, that is; "you missed a minus sign here" doesn't need any further elaboration).

But I know many of the other tutors just checked key points in the working out. They got through their assignment loads far faster than I did, but I hope I provided a better service to my students.

I had a couple of cases where students used a non-obvious methodology, very different to the one on the answer sheet, but managed nevertheless to correctly derive the answer. They got full marks for that, of course. The answer sheets were often buggy, too - this helped me to catch cheaters more than once.

I think the funniest one I got was for one question where the answer sheet was fairly wrong. Two students in my group (whom I'd previously noted as copying occasional questions from each other) had copied the answer sheet

exactly. Identical diagram, same choice of variable names, every error in the logic faithfully reproduced. Their reward (as sanctioned by the official policy at the time) was a mark of 0 for the whole assignment.