I don't think that's what imaginary numbers stem from. If I recall, they came from the much more specific question of "what if I apply the formula for roots of a cubic polynomial, even when that formula produces some nonsensical intermediate values that end up canceling each other out?".
It's less about "what if we do this?" and more about "we have this method, and it seems to work even in cases where it's not obvious that it should work".
There must be a study of this; failed attempts of taxonomy.
For what it's worth, you can avoid "imaginary" numbers entirely if you want -- you can do arithmetic on real polynomials modulo x^2+1, or equivalently on 2x2 real matrices of a special form.