The following is, I assume, a known fact---however, the only proof I know uses the axiom of choice (then a theorem about well-ordered sets as well as transfinite recursion). I was wondering if anyone knew a more elementary solution.
Let $a_0$, $a_1$, ..., $a_n$ be a sequence of real numbers, $a_n \neq 0$. Then the functional equation $$ a_0x + a_1f(x) + a_2 f^{(2)}(x) + ... + a_n f^{(n)}(x) = 0 $$ always has a solution.
Here $f^{(n)}$ denotes function $f(f(...(x)...))$, that is, the $n$-fold composition of function $f$ with itself. And "function" here means an absolutely arbitrary mapping from $\mathbb R$ to $\mathbb R$---it does not have to be continuous or bijective etc.
Of course, if polynomial $P(t) = \sum_{k=0}^n a_kt^k$ has a real root $r$, then function $f(x) = rx$ is such a solution. However, when all roots are complex, the matter becomes quite non-trivial (but solving this problem for an arbitrary quadratic polynomial $Q(t)$ would be enough, as any functional root of $Q$ would also be a functional root of any polynomial divisible by $Q$).