I’m trying to figure out the fundamental limits to scaling quantum error correction beyond a few hundred logical qubits. When I worked out the overhead for a distance-7 surface code on a 2D grid with 0.1% gate errors, the physical qubit count exploded after I added in the routing and measurement overhead. Not sure if decoding latency or the long-range connections end up being the bigger problem, or whether something else is the real constraint when you push to larger systems. What actually sets the scaling ceiling here?