Paths are driven by the client and unchecked, it's not hard to do. However first you're better off taking a serious look at your threat model, which you've not mentioned at all for anyone to help you qualify it and any potential solutions therein, before turning random knobs in whatever anonymous overlays you're using.
Likely more useful than hop count would be analysis of the real world node network, with path selection leanings and choices being made therein. But again, no one seems to be taking up and working that project. Sybils all the way down...