I'm pretty sure that the destinations used for drawing unit paths are
not recalculated as often as the destinations used for deciding where
to actually go. I know this by observing that they are so often wrong
for periods of time much longer than 5 turns (perhaps 100s of turns?).
I suspect the pathfinding destinations are being copied and cached
somewhere and this copy is not being updated.
--
Joe
P.S. Anway, it is clear that the pathfinding destinations must be
computed regardless of whether they are being displayed (otherwise how
could globules move at all?), so there can not be any extra expense
for display purposes in computing them.
The problem is that the globules never find a "destination", atleast when they are going to ressources. They merely roll down a path, a "hill" so to speak. You might not know where you will end up at the top, but if you start at the same point, your going to roll down and arround following a precise path untill you reach the bottom (the destination). There are many places on the bottom you could have ended up, but your starting point meant you would land at this one. Our pathfinding is the same way, the paths to all ressources are computed at the same time, where each position is given a value based on the distance to its nearest ressource. It doesn't know what that nearest ressource is, but it knows how far it is away. Glob2 just follow the path to the ressource by making sure they always get closer to the ressource (they make sure that they decrease the distance with every movement). We can only estimate where they will end up.