There is some good info here: stats.stackexchange.com/questions/135052/…

Not sure about advantages as such, but if you have a 2x2x8 set of data and want to test the association across the groups it is the correct one to be using. The other approach could be log-linear modelling (aka Poisson regression) where you set and interpret model parameters for the associations you are interested in.

Probably depends on how non-linear and your sample size. Bergtold et al. has a good article on this tandfonline.com/doi/abs/10.1080/02664763.2017.1282441