Well, they do take one less step to move from x,y,z to x+1,y,z+1. There's a reason why the stairs test was larger: had to add two straight tiles to the non-stair path.
But always good to test something before you plan your whole fort around it :v (I learned this lesson the hard way.)
I did a further test with this setup
|
|
Since it occurred to me that the previous test might suffer from acting like 1-way paths (noted on the wiki). Both ways were used again, right on movement, left on return. Ditto for this setup for exclusively up/down movements:
|
|
|
|
Here I also tested mining, since I recall it was supposed to have different job pathing rules than regular jobs, and the results were mirrors of each other. For building, on the left the dwarf took southmost stairs, and on the right first stairs they encountered. For mining, on the left they took first stairs, and on the right northmost stairs. But I tought a bit, and realized these result also make sense if mining and building are the same, but idle pathing is different, since in mining case they finished right second and returned with no job, while in building case they finished left second and returned with no job, and in both cases they took the first available stairs to the dwarf when taking a new job and preferred lowest available z-level for idle pathing.
Of course, that means the pathing algoritm did pick last suitable stair for the job, which kinda supports your claim.
But ultimately, the important part is fps effect (though the above tests are still useful for setting up pressure plates). I don't have any 3x3 staircase forts handy, but someone who has them could relatively easily test the effects of reduced number of staircases with bit of savescumming and gui/liquids. Set up bunch of large jobs that force taking it, record three times it took to pass a month (via dfhack to record the time, unpause at start and pause the game at that exact moment), then do it again but with some part of staircase replaced by obsidian walls. I also recall observations of some shortcut staircases hampering the fps due providing alternate longer path.
From efficiency point of view, I was bigger about cube forts 'till I realized that stockpile to workshop dominates the calcs.