The basic gist of it is that the AI will follow a path until it reaches it's conclusion. Turret and subsystem paths are relatively simple, as they basically just provide an approach vector for AI attacking the associated turret/subsystem. Fighterbay paths are one step more complex, in that they have to be called "$BayXX" to work IIRC. Lastly, there are docking paths. I will state that all paths work with one vertex, and work sufficiently with 2. The only reasons to add more are to either add some manuvering to the launch/land sequence on your hanger-equiped ship, or to make docking work. Almost all dock paths have 4 points, because they are used for something like approach from 1-2, rotate into correct alignment from 2-3, move into position from points 3-4. Once the dock path is run, the docking ship will simply follow the shortest line between the last vertex point and the first associated dockpoint. That's the arcane science of pathing, pure and simple.
Simply put, having too many or too few points in a given path should not crash the game (ok, 0 points can, but 1+ should not) but can cause some really screwball behavior when fighters try to run them.
Also, for turret and subsystem paths, the autopath function on Aurora is usually perfectly acceptable. It's the other paths that you have to worry about.
EDIT: as aldo said, it really depends on how complex the geometry is around a specific path as to how many points you need. I've used anywhere from 2 to 6 points in pathing my caps, but I generally stick to 4 because it usually happens to work out that way. The biggest catch is figuring out which way the paths need to be ordered, as having them backwards creates some, well, interesting results.
