I read somewhere on this site that for a branch cut between $(-\infty, 0]$, the range of values for $\arg(z)$ is $[-\pi,\pi]$, while for a branch cut between $[0,\infty)$, the range of values is $[0,2\pi)$. Why is this so? Is this just a definition, or is there some mathematical reasoning to it? Does this apply to branch cuts along the imaginary axis? Also, for a branch cut that crosses the origin, such as $[-1,1]$ for $\sqrt{z^2-1}$, what should the approach be?
Edit: the question on stackexchange I was referring to is this one.

The ${\rm arg}$ function is multiple valued on $\dot{\mathbb C}$. In order to obtain a well defined and continuous (even smooth) real valued function on a slightly restricted domain one then makes the branch cuts. If the cut is made along the negative real axis one then defines $${\rm Arg}(z):={\rm the}\bigl({\rm arg}(z)\>\cap\>]{-\pi},\pi[\>\bigr)$$ as principal value of the argument, and if the cut is made along the positive real axis one can take $$\phi(z):={\rm the}\bigl({\rm arg}(z)\>\cap\>]0,2\pi[\>\bigr)$$ as smooth real representant of the argument. In both cases the chosen function jumps by $2\pi$ when you cross the cut. It would be pointless to make a cut along the negative real axis and then choosing $\>]0,2\pi[\>$ as range of the argument.
The cut along $[-1,1]$ is of no help for the argument function, but in other cases it is. The function $$f(z):=\sqrt{z^2-1}\qquad(?)$$ is double valued for all $z\ne\pm1$. But if you make a cut along $[-1,1]$ you can single out the branch that is $\>= +z+o(1)$ for $|z|\to\infty $, and will never get into trouble, i.e., obtain a single valued function $f$ defined on all of ${\mathbb C}\setminus[-1,1]$.