[0] http://research.microsoft.com/en-us/um/people/lamport/pubs/l...
[1] http://research.microsoft.com/en-us/um/people/lamport/pubs/p...
[2] https://news.ycombinator.com/item?id=10689391
[edit: formatting]
Like that time someone solved an unsolved problem or two because it was on his homework:
Anyway, whether something counts as a proof is a matter of whether you (logically) accept it as irrefutable evidence. As with any genre, you have to consider your audience. I hear gauge theorists are content to have as a proof a couple of references to a couple papers which have techniques which apply, with minor alteration, to the current problem.
That said, I wouldn't have minded if Zagier actually demonstrated it was an involution with exactly one fixed point!
For example, proof that real numbers are not countable:
If r_n is the n-th real number in [0,1) and n{k} is kth digit of n in base 3, then consider sum(3 - r_j{j})/3^j (QED)