This isn't very deep. But it might suggest better results:
I will assume $1 < x <y.$
I will relax the condition on $z$ slightly to $z>1$ and $z\mid \gcd(x-1,y-1).$
It would be enough to solve this for $z$ a prime or prime-power and then combine results.
In addition, I will sometimes ignore the condition $\gcd(x,y)=1$.
When I wrote this I was assuming that $x$ was a factor of one of the $n-i$, for example $z=5$, $x=21$, $k=2$ has solutions
- $(n-1)n=42\cdot 43=21 \cdot 86$ with $5=\gcd(20,85)$
- $n(n-1)=273\cdot 272=21\cdot 3536$ with $5=\gcd(20,3535)$
- $(n-1)n=77\cdot 78=(7 \cdot 11)\cdot(3 \cdot 26)=21\cdot 286$ with $5=\gcd(20,285)$.
I can describe all the cases of type 1 and type 2. But the following does not consider type 3.
I will suggest that for each fixed $k$ the possible $z$ (subject to $x$ being a divisor of one of the $n-i$) can be described and then we can take $x=zt+1>1$ arbitrary and describe exactly which $y$ are then possible. These will be a union of arithmetic progressions. I describe this in any detail only for $k=2$. Larger $k$ should be similar but with more subcases.
Consider first the case $k=2$. There are three cases of which I only consider the first two:
- $n-1=ax$
- $n=ax$
- $n-1=ax_1$, $n=bx_2$ with $x_1,x_2>1$ and $x=x_1x_2$.
In the case 1) of $n-1=ax$ the condition on $z$ turns out to be that $r^2+r-1$ has solutions $\bmod z$. This is equivalent to saying that either $z=\prod_1^j p _i^{e_1}$ or $z=5\prod_1^j p_i^{e_i}$ where each $p_i \equiv \pm1 \bmod 5$. Then there are $2^j$ solutions $\bmod z$.
We will now see these solutions when $z=5$ are exactly the following when $t\ge 1$, $u \ge 0$. (Note that $r^2+r-1=(r-2)(r-2) \bmod 5$.)
- $z=5$
- $n=(5t+1)(5u+2)+1$
- $x=5t+1$
- $y=\left( 5\,u+2 \right) \left( 25\,tu+10\,t+5\,u+3 \right)$.
To get that $\gcd(x,y)=1$ we must also require $\gcd(u,2t)=1$.
and
The solutions when $z=11$ are exactly the following when $t\ge 1$, $u \ge 0$ and $r=3$ or $r=7$. (Note that $r^2+r-1)=(r-3)(r-7) \bmod 11$.)
- $z=11$
- $n=(11t+1)(11u+r)+1$ for $r=3$ or $r=7$
- $x=11t+1$
- $y=\left( 11\,u+r \right) \left( 11\,rt+121\,tu+r+11\,u+1 \right)$.
To get that $\gcd(x,y)=1$ we must again require $\gcd(u,tr)=1$.
In general:
For
- any $z$ of the form above
- any $x=zt+1>1$ and
- any $a=zu+r$ (for one of the possible values of $r$)
take $n=ax+1=a(zt+1)+1$. Then $$(n-1)n=ax
\cdot (ax+1)=(zt+1)\cdot \bigl((zt+1)a^2+a\bigr)=x\cdot y.$$
The extra condition $\gcd(x,y)=1$ is equivalent to $\gcd(zt+1,zu+r)=1$ which is equivalent to $\gcd(u,tr)=1$.
We arranged things so that $z\mid x-1$ and wanted that $z\mid y-1$ as well. That is
$$z \mid a^2(zt+1)+a-1\ \ \ \ \text{i.e.}\ \ \ z\mid(a^2+a-1).$$
But we chose $a$ exactly so that would be true.
In the case $n=ax$ we need that $s^2-s-1 \bmod z$ has solutions. The feasible $z$ are the same, and the solutions are exactly $z-r$ where $r$ is a solution for $r^2+r-1=0 \bmod z$.
When we go to $k=3$ we need to consider the subcases $n=a(zt+1)+2$, $n=a(zt+1)+1$ and $n=a(zt+1)$. In each case $a$ must satisfy a cubic equation $\bmod z$. These are less elementary. But the concept remains the same.
It is a necessary, but not sufficient, condition that all prime divisors of $z$ are greater than $k$, i.e. $\gcd(z,k!)=1$. This is because $k! \mid n^{\underline k}=xy$ and $\gcd(z,xy)=1$.
$$
. For example, set braces{}
disappear from your TeX if you just cut and paste. I have edited accordingly. $\endgroup$