A* pathfinding around circular obstacles produces two kinds of route segment: straight lines between obstacles, and arcs that hug an obstacle’s boundary exactly as far as the geometry forces. The algorithm can prove this: for a given field of obstacles it finds the route with the least possible arc, and a shorter one provably does not exist. Every foot of hugging in the optimal path is demanded by geometry, not chosen by anyone.
Housing permitting looks structurally similar and is not the same kind of problem. A comparison of build rates in Austin and San Francisco found the same money, the same technology, the same number of qualified builders, and an eightfold gap in homes actually permitted per resident. The gap traces to a permitting path that hugs a checkpoint added in the 1970s, another from the 1990s, another from the 2000s, each locally reasonable when it went in, none of them re-evaluated against what was stacked before or after it.
Permitting constraints are not points and arcs with one number to minimize, so there is no version of A* to run on them. But A* still marks what a genuine check would look like: it re-solves the whole obstacle field every time a new obstacle appears, and proves the resulting detour is the smallest available. Permitting has no equivalent moment. Each checkpoint gets reviewed once, against the obstacles that existed when it was added, and never again against the ones added since. The 1970s environmental study has never been checked against the 2000s comment period to see whether both are still doing distinct work. The rules might already be minimal. Nobody has looked at them together to find out.