Definitions are just between you and me, so we both know what we talk about. Else i make a big mess of it all.
Yes a cell size of 8 and a -1 border, with a big tilemap, is CPU intense.
You do realise that this is a new question eh ? The thing is, you wanted obstacle cells that exactly represent your solids tilemap. You dont want unwanted cells marked as obstacles. I think i answered that question for you.
Personally, i dont see any use in this. Because, see. The origin of an object moves over the found path, the pathfinder is totally dumb about the size of that object. Since it only keeps the origin on path, it will drive objects into the solids. The origin will always keep out the solids, ofcours.
Now there are many kinda-sorta solutions on the forum to deal with badly calculated paths due badly set cell-size and border-size. Me myself has tried and experimented a lot. And the only thing that helped me out, is to calculate good paths.
A good path is a path that stays half the size of the objects away from the solids. Or. The collisions cells have cells marked as obstacle around the solids. The opposite of what you wanted.
Or in you case: cell size = tile size of the tilemap holding the solids. Border size = half the size of the moving sprite , and to be experimented. (no other choice) Check the cells by making the cells visible in a obstacle tilemap. If it does not find a path, you need to change the solids.
Again, if you look at my 'blah example', did you see that you have to WAIT 1 tick. The obstacle cells are READY the NEXT tick. Even when starting up the game.
Additional. There is a difference between the graphical representation of a wall and the way/form/whatever it should act as an obstacle. You better have 1 tilemap for graphical representation and 1 tilemap + solid sprites (all invisible) to mark obstacles. Because, see, someone/thing got to do all the work, YOU or the players (victims) PC.
I am not harsh, i just type that way in bad english.