Is it immediately apparent that the solution to the system of equations,
$$\begin{aligned} x_1^2 &= x_2+2\\ x_2^2 &= x_3+2\\ x_3^2 &= x_4+2\\ &\vdots\\ x_n^2 &= x_1+2\\ \end{aligned}\tag{1}$$
can be given by the roots of unity? Specifically,
$$x =\frac{y_k^2+1}{y_k}\tag{2a}$$
where the $y_k$ are,
$$\begin{aligned} y_k &= \exp\Big(\frac{2\pi i k}{2^n-1}\Big),\; k = 0\dots 2^{n-1}-1\\ y_k &= \exp\Big(\frac{2\pi i k}{2^n+1}\Big),\; k = 1\dots 2^{n-1}\\ \end{aligned}\tag{2b}$$
Example. Let $n=4$. Then (1) is equivalent to,
$$x = (((x^2-2)^2-2)^2-2)^2-2\tag{3}$$
Expanded out, (3) is a $2^4=16$-deg polynomial and its 16 roots are given by (2) where,
$$y_k = \exp\Big(\frac{2\pi i k}{15}\Big),\; k = 0\dots 7$$ $$y_k = \exp\Big(\frac{2\pi i k}{17}\Big),\; k = 1\dots 8$$
Ramanujan considered the system (1) for $n=3,4$ in the general case and also as nested radicals. For $x = (((x^2-a)^2-a)^2-a)^2-a$, see this related post. (Interestingly, $n=5$ in the general case is no longer completely solvable in radicals.)
Question:
I observed (2) empirically. How do we prove from first principles that this is indeed the solution?