Here is a theorem from Axler's book "Linear Algebra Done Right" that should be helpful.
Let $V$ be a finite dimensional real inner product space, and $T$ a linear operator on $V$. Then $T$ is normal if and only if there is an orthonormal basis of $V$ with respect to which the matrix of $T$ has block diagonal form with each block either $1\times 1$ or $2\times 2$. The $2 \times 2$ blocks are of the form
$$\bigl(\begin{smallmatrix}
a & -b\\
b & a
\end{smallmatrix}\bigr)$$
with $b > 0$.
Since the minimal polynomial of $T$ is an irreducible quadratic, it has no real eigenvalues. Thus there are no $1\times 1$ blocks in the above decomposition. Using this, you can see exactly the form of $T$ and $T^*$.
The eigenvalues of the matrix $$\bigl(\begin{smallmatrix}
a & -b\\
b & a
\end{smallmatrix}\bigr)$$
are $a \pm bi$. So if there were more than one type of $2\times 2$ matrix showing up in the decomposition, then we would have more than two eigenvalues, and so our minimal polynomial would not be a quadratic. Therefore our decomposition is just a bunch of copies of
$$\bigl(\begin{smallmatrix}
a & -b\\
b & a
\end{smallmatrix}\bigr)$$
along the diagonal.
So lets find a polynomial that will take the transpose of this $2\times 2$. Since taking the transpose will make the upper-right entry $b$, I will first multiply the matrix by $-1$. This gives
$$\bigl(\begin{smallmatrix}
-a & b\\
-b & -a
\end{smallmatrix}\bigr).$$
But now the main diagonal is bad. We can fix it by adding $2a$ to the main diagonal. We can accomplish this by adding the matrix $2a I$.
So $$\bigl(\begin{smallmatrix}
a & -b\\
b & a
\end{smallmatrix}\bigr)^* = \bigl(\begin{smallmatrix}
a & -b\\
b & a
\end{smallmatrix}\bigr)^{tr} = -\bigl(\begin{smallmatrix}
a & -b\\
b & a
\end{smallmatrix}\bigr) + 2aI.$$
Therefore the polynomial we want is $f(x) = -x + 2a$.
You can now argue that this polynomial works for $T$.