Also consider that fact that human IMO contestants have little training in university-level mathematics. The problem setters try to choose problems where knowledge of advanced mathematics doesn't help, in order to produce a level playing field which only measures raw problem-solving ability. But I suspect that postgraduate-level mathematics will nevertheless be useful in programming the AI.
I don't think this is true, as in my experience many of the contestants have already cultured a background in calculus but refrain from using it as the problems are usually designed to actively discourage its use.
Terrence Tao definitely says that doing university math changed his approach to these problems. See https://terrytao.wordpress.com/books/solving-mathematical-pr...
Calculus.
Uni adds a much more axiomatic and formal approach.
The answer turns out to be pretty boring: As long as humans are still better than machine, humans can provide algorithmic insights and be a source of cheap labels.
So I am pretty pessimistic about this particular task. There are only a small handful of humans in the world who are qualified to help! (according to IMO scores: https://www.imo-official.org/hall.aspx)