When you compute $\partial x/\partial r$, it's understood that you're supposed to keep $\theta$ constant, but when you compute $\partial r/\partial y$, you're supposed to keep $x$ constant, which is something completely different. In view of this, there shouldn't be any reason to expect that $\partial x/\partial r = 1/(\partial r/\partial x)$.
What is true, however, is that the Jacobian matrices are each other's inverses:
$$
\begin{pmatrix}
\partial x/\partial r & \partial x/\partial \theta
\\
\partial y/\partial r & \partial y/\partial \theta
\end{pmatrix}
=
\begin{pmatrix}
\partial r/\partial x & \partial r/\partial y
\\
\partial \theta/\partial x & \partial \theta/\partial y
\end{pmatrix}^{-1}
.
$$