I just had a chat with the development team to get some insights into what's happening behind the scenes. Basically, there are two steps:
1 - Preparation of all assigned segments, regardless of if they are used in the scene or not.
2 - Processing the segments that are actually used.
Step 1 does apply to all proxies. But the performance impact of this step should be minimal so you should be able to assign all the segments you need without concerns about performance.
That said, there is a limitation with Corona proxies, over which I am afraid we don't have control, which means that RailClone does poll the whole mesh in step 1. So for Corona, many unused proxies can cause a performance hit. We're in conversation with the Corona team about some changes required on their side to improve that, but I don't have a timescale at this point.
I hope that answers your question and many thanks again.