From the $1995$ All-Russian Olympiad, $9^{th}$ Graders, Final Round:
Is it possible for the equation $f(g(h(x)))=0$, where $f, g$ and $h$ are quadratic functions, to have solutions $x=1,2,\dots,8$?
I am going to post my own solution so that people can check if it's valid or the most elegant.
Essentially, $h$ has to map $8$ values into $4$, $g$ maps $4$ values into $2$, and $f$ maps $2$ values into the single number $0$
Now I'll try to make $f(g(h(x)))$ symmetric around the $y$-axis for ease of calculation.
If it's possible to have $f(g(h(x)))=(x-1)(x-2)...(x-8)$ it should also be possible to have
$f_1(g_1(h_1(x)))=(x-3.5)(x-2.5)(x-1.5)(x-0.5)(x+0.5)(x+1.5)(x+2.5)(x+3.5)$
which is simply $f(g(h(x+4.5)$ but $h(x+4.5)$ is also a quadratic in $x$
And consequently $f_2(g_2(h_2(x)))=f_1(g_1(h_1(x/2)))=\dfrac {1} {2^8} (x-7)(x-5)...(x+5)(x+7)$
Again $h_1(x/2)$ is a quadratic in $x$
If $h_2$ is to map $7,5,...,-5,-7$ to $4$ values, it must be symmetric around the y-axis and must be of the form $x^2+a$
But the resulting $4$ values which $g$ has to map is $1+a, 9+a, 25+a, 49+a$, and to map them into $2$ values, points $1+a$ and $49+a$ have to be equidistant from some $c$, and so does $9+a$ and $25+a$ (but from the same $c$). Clearly there is no such $c$.