You only get NaN for division by zero if you divide 0 by 0 in IEEE floating point. For X/0 with X ≠ 0, you get sign(X)•Inf.
And for real numbers, X/0 has to be left undefined (for all real X) or else the remaining field axioms would allow you to derive yourself into contradictions. (And this extends to complex numbers too.)
As a not good at math person, advanced math sounds hella fake sometimes. Like almost a parody of math. I accept that it's real but part of me will always think it's some inside joke or a secret society cult for the lamest god.
Matrix representations in general, if that counts?
Complex numbers, polynomials, the derivative operator, spinors etc. they're all matrices. Numbers are just shorthand labels for certain classes of matrices, fight me.
I'm just being silly, but I mean that if everything can be represented as a matrix then there's a point of view where things like complex numbers are just "names" of specific matrices and the rules that apply to those "names" are just derived from the relevant matrix operations.
Essentially I'm saying that the normal form is an abstract short hand notation of the matrix representation. The matrices are of course significantly harder and more confusing to work with, but in some cases the richness of that structure is very beautiful and insightful.
(I'm particularly in love with the fact one can derive spinors and their transforms purely from the spacetime/Lorentz transforms. It's a really satisfying exercise and it's some beautiful algebra/group theory.)