After looking at many sums of squares, this is something I am hypothesizing. I'd love to know if there is any work related to this, if it is easy to solve or even if there exist counterexamples.
Conjecture: Let $p$ be some prime number where $p \equiv 1 \bmod 4$. Then, for $n \in \mathbb{N}$ and $n \geq 1$, $2p^{2n}$ will give $n$ unique pairs $a,b \in \mathbb{N}$ such that:
\begin{align*} a^2 + b^2 &= 2p^{2n} \end{align*}
In the image below I have provided 3 different examples along with the $a,b$ sums of squares for more details.
Ex 1: $p=617$ and $n=1$.
Ex 2: $p = 13$ and $n=3$.
Ex 3: $p = 5$ and $n=7$.
There’s an exact formula for the number of ways to express given integer as a sum of two squares(by the theory of modular forms). You can use this to prove your statement.