Estoy tratando de trabajar fuera de la gran $n$ asymptotics de $$S_n = \prod_{x=1}^{\lceil\frac{n}{\log_2{n} }\rceil} \left(\frac{1}{\sqrt{n}} + x\left(\frac{1}{n}-\frac{2}{n^\frac{3}{2}} \right)\right) .$$
Here is my attempt so far $$\prod_{x=1}^{k} (A + Bx) = \frac{B^k \Gamma(k+1+A/B)}{\Gamma(1+A/B)}.$$
In our case $A/B \approx \sqrt{n}$ and $\left(\frac{1}{n}-\frac{2}{n^\frac{3}{2}} \right) \approx \frac{1}{n}$. Therefore
$$S_n \approx \frac{\frac{1}{n}^{\frac{n}{\log_2{n}}} \left(\frac{n}{\log_2{n}} + \sqrt{n}\right)!}{\sqrt{n}!}$$
I am not really sure where to go from here, if I haven't already taken an approximation too far.
I tried taking logs and defining in maple
f:=(x,n)-> log(1/sqrt(n)+x*(1/n-2/n^(3/2)))
Now if you do
plot(-(sum(f(x, n), x = 1 .. n/log(n))), n = 10 .. 100)
you get what looks like a linear function of $n$. However if you do
limit(-(sum(f(x, n), x = 1 .. n/log(n)))/n, n = infinity)
you get $0$.