Consider positive numbers $ x_0,\ldots, x_{n-1}$, and extend the sequence cyclicly by setting $ x_i=x_{i\,{\rm mod}\,n}$. Now, for $1\le k\le n$ we define $ f_k(t_0,t_1,\ldots, t_{k-1})=\prod_{j=0}^{k-1}t_j$ and we set $$F_{k,n}(x_0,x_1,\ldots,x_{n-1})=\sum_{j=0}^{n-1}f_k(x_j,x_{j+1},\ldots,x_{j+k-1})$$ I am considering $$M_{n,k}=\sup_{x_0+\cdots+x_{n-1}=1}F_{k,n}(x_0,x_1,\ldots,x_{n-1})$$ Testing sequences of the form $x_0=\cdots=x_{k-1}=1/k$ and $x_0=\cdots=x_{n-1}=1/n$ Shows that $$M_{n,k}\ge 1/\min( n^{k-1},k^k).\tag{$*$}$$ It is not hard, to show that we have $M_{n,k}=k^{-k}$ when $1\le k\le n/2$, so we have equality in $(*)$ in this case. Also, it is easy to prove equality in $(*)$ when $k=n,n-1$.
The remaining cases seem intractable to me.
My question is: Does equality in $(*)$ hold when $\frac{n}{2}< k\le n-2$ ?