Motivated by Kevin Liu's recent question, here I pose the following conjecture based on my numerical computation.
Conjecture. Let $m>1$ and $n>1$ be integers. Let $\delta\in\{0,1\}$ and let $\zeta$ be a primitive $(m(n-\delta)-(-1)^{\delta})$-th root of unity. Then, for the sum $$S:=\sum_{k=1}^{n-1}\left(\frac{\zeta^k}{1+\zeta^{km}}-(-1)^{n-k+\delta}\frac{\zeta^k}{1-\zeta^{km}}\right),$$ its real part is $$\text{Re}(S)=(-1)^{n-1}\left\lfloor \frac n2\right\rfloor.$$
The case $\delta=1$ of the conjecture might be handled by the method of Fedor Petrov used in his solution of Liu's question, but the case $\delta=0$ looks challenging. Your comments are welcome!