Hey Mieszko, thanks for the info.
So I’ve looked at the code and I think I see what you mean. String pulling seems to be on by default, and it makes sense since if it wasn’t done at all then pathing routes would surely look strange. But I wonder in that case why cost calculation is done before string pulling, rather than afterwards in order to get more accurate costs?
Changing to the ChunkyMonotone partitioning allows me to reduce polygon size and so lessen the effect, but it seems very inefficient to create lots more polygons just for this purpose. Am I missing a better solution? My environments will generally be very flat and mostly have a single area cost - I don’t quite understand in what way you mean playing with area cost could help?
Cheers.