I've been testing diferent costs for docking tile occupancy, to experiment with their intended purpose of spreading ships to diferent docking tiles, and I realise 300 may be too little. A value around 1400-1500 resulted in a better spread.

1500 cost: 12 ships, 12 docking tiles, ships occuppied 12/12

300 cost: 12 ships, 12 docking tiles, ships occuppied 6/12

Perhaps if this PR is accepted, docking tile penalty cost could also be adjusted for a better spreading of ships.

PS: These tests were done with the #8021 fix already in.

This is just a thought, but how is distance to the station calculated (for adding the penalties)? Does it take into account the setting for "station spread"?
I ask this because it could look mighty strange if a ship were to sail past an empty dock only to turn around and sail back to it again because the penalty wasn't added soon enough.

I am not convinced at all that this is a proper way to solve ship path-finding. It might work for the cases you tried, but it reads to me many more cases will start failing horribly. Basically, you are chunking the path-finders, which is good for performance, but gives many many side-effects most of the time.

Additionally, the way you describe your patch feels to me that you don't really understand what is going on, but that this just happened to work. This too is a reason I am not convinced.

I would need an opinion of someone with in-depth knowledge of our path-finders to convince me otherwise. Or someone who really did his research here, and showed many different cases to proof that common use-cases don't break now.

In general, ship path-finding is hard, because of the open seas. It will always be slow, unless someone comes up with a complete novel idea how to solve it.