The key parts that I found I had to spend time on in my program were:

1. Calculating distances. I had to get rather closer to geodesy than I normally do. Map projections and so on. My approach should work for reasonably small radii (eg my 20km).
2. Data cleaning. The program has some ad hoc exception tables in it. The data exception system in my program is ad-hoc, bespoke, abstract, and undocumented. Sorry about that...