There is no systematic approach to considering curves along which the limit might assume distinct values. This example is a very particular one: among all lines passing through the origin it is zero, but among any parabola it takes another value.
This is one of those things where experience is handy: solving many problems and tackling with different approaches is what guides us.
Perhaps the intuition for taking a parabola in this case is noticing that the exponent of $y$ is exactly two times the $x$. You can construct many similar ones in this way:
$$f(x,y) = \frac{7x^3 y}{x^6 + y^2}, \quad f(x,y) = \frac{13x^5y}{x^{10}+y^2},$$
just to create a few. In these cases you'll find the same: along lines it is zero, along curves such as $y=x^3$ and $y=x^5$ each will have a different limit.