Oversampling is the process of creating or receiving a signal in a higher sample rate than the system rate in order to be able to downsample this. The advantage of this approach is to minimize aliasing artifacts.
Notice that if you have two machines connected to each other, with oversampling enabled on both, the signal is still downsampled between each of these machines. So if you enable oversampling on all machines, you would actually be better off running the entire song at the double sample rate. This would both maintain the extra frequency headroom and use less CPU, as there would not be multiple downsampling steps.