Continued work on the router. I have settled on what I consider a good solution to the problem of finding interesting quest routes.
1. Points in a circular band of variable width are collected.
2. They are filtered on undesirable characteristics, like initialization state and proximity to the origin or other special tiles.
3. They are ranked on factors of likely traversal paths, like the state and type of tiles on the map between the two points.
By following only the orthogonal or diagonal paths to maximize time spent on the diagonal, points can be optimally navigated using just two directions. The traversal paths mentioned in the last step are based on two possible paths: Following to completion either the diagonal or orthogonal direction, then the other (if necessary). This technique mimicks the discrete ordinals quest compass telling the player roughly which way to go.
Normally, this would get less adequate at greater distances. The number of possible “interior” optimal path variants would probably have a more navigable route. A player could still “wiggle” between the two semi-ordinal directions indicated to him or her, and so adopt a path not tailored to the generating quest’s specifications. On the other hand this is a reasonable invitation for strategizing. At sane distances, this routing method offers a good directional guideline, and minimizes excess calculations.