When calculating the matrix element for (let's say) $e^+e^- \to \mu^+\mu^-$ we have to average over initial spins and sum over final spins. I understand the motivation of this, but when the calculation is done, the sum is done for 2 cases: spin up and spin down, so you have to add 8 terms (2 for each particle) and divide by 4 (the 2 spins of the 2 initial incoming particles).
Why is this summation enough? Shouldn't one integrate over all possible values of spin? It is not like the particle will come with either spin up or spin down on a given axis, they can be a linear combination of these.
I also noticed that in the massless limit they do the same approach using $e_R$ and $e_L$ separately and then divide by 4? Why is this enough? You get the same result doing this as you would get by integration or am I missing something?