Sorgenfrey line demonstrates how normality can be not preserved when "squared."
Is there an example for a normal space $X$ for each of?:
$X^2$ is normal, but $X^3$ is not
$X^2$ and $X^3$ are normal, but $X^4$ is not
$X^k$ is normal for $2 \leq k < n$, but $X^n$ is not
$X^n$ is normal for $n \in \mathbb N$, but $X^\omega$ is not
$X^n$ is normal for $n \in \mathbb N$, but $X^\omega$ is not, in box topology
For the first four questions, the answer is "yes", and I believe was first established in the paper
Therein the following theorem is proved:
The answer to the last question is also "yes", and an example was essentially established in
(I.e., the box product of $\omega$ copies of the Tychonoff product $2^{\omega_2} = \{ 0,1 \}^{\omega_2}$ is not normal.) Clearly $2^{\omega_2}$ is compact Hausdorff, hence normal, and each finite Tychonoff product $( 2^{\omega_2} )^n$ is homeomorphic to $2^{\omega_2}$.