Perhaps I'm a bit rusty on my number theory, but I have been struggling to figure out how to determine what positive integers $x$ and $y$ satisfy $$ x-y \text{ divides } x^2 + y^2 $$
Any thoughts? I have tried looking at a remainder after polynomial division, adding/subtracting multiples of the divisor, and factoring out a gcd term to see if that helped.
Any input would be appreciated.
Edit: If it is any help in answering the question, this is in order to help figure out how to prove the following:

To solve $x^2 + y^2 = k(x-y)$, you basically write it as $(2x-k)^2 + (2y+k)^2 = 2k^2$, and then use facts about the solutions to $a^2+b^2 = N$ alongside parity arguments to count the number of solutions.
Observations towards a solution:
The condition $x+y > 4$ is somewhat forced / complicates things. For now, we remove that condition and add in the 4 solutions $(1, 2), (1, 3), (2, 1), (3, 1) $.
For now, we also remove the condition that $x, y$ are positive, and account for that later.
If $ \frac{ x^2 + y^2 } { x-y } = \frac{ 1995}{k}$, then $\frac{ (kx)^2 + (ky)^2 } { kx-ky} = 1995$.
To solve $x^2 + y^2 = 1995 (x-y)$, it is the same as solving $(2x-1995)^2 + (2y+1995)^2 = 2\times 1995^2 $.
Now we do number theory, especially using facts about sums of squares
Now, we add back the condition that $x, y> 0$.
For each of the two solutions, $\gcd(x,y) = 399$ and the number of (positive) divisors of 399 is $\sigma_0(399) = 8$. As mentioned earlier, each divisor is a choice of $k$ that gives a unique solution. For the 8 additional negative choices of $k$, we get negative values for $x$ and $y$, which we turn into a valid solution by making both values positive and interchanging them ($(-x,-y)$ becomes $(y,x)$). These solutions are not counted for by the positive $k$ solutions since they give different values for the divisor of 1995 $\frac{x^2+y^2}{x-y}$.
Therefore, the final number of solutions is $16 + 16 - 4 = $ 28