The structural elements of the underground and above ground pedestrian network does two things, both pernicious to pedestrian activity on the ground plane. First, they remove people from the street. This can only work in hyper density and even in that case, there is no evidence from other cities that is an ideal condition as one generally makes the other less than comfortable. Second, the structural elements (and walls of T-Giving square) limit line of sight and (perceived) safety. When we can’t see what is on the other side of something, particularly when there aren’t many people around to “police” (social contract kind of thing) an area, people don’t like being there. This is well documented by William Whyte, Jan Gehl, and Kevin Lynch.
Also, I’m not convinced we should be talking about pedestrian only networks yet. In one way, mostly theoretical, they can be good for overall interconnectivity. But again, without lacking suitable density, they just dilute the movement (energy) from other streets, all of which become less than ideal for ground floor businesses which survive off movement and the visability afforded by movement. And by movement, I mean traffic, and by traffic I mean all modes of transportation.
I should add that between the narrow sidewalks that lack any buffer from traffic that is designed to move quickly and the N-S pedestrian route you describe and its incomprehensible nature as well as uncomfortable experience driving people from one to the other, neither being comfortable and amenable. In the end, people just don't walk. No people walking, no incentive for businesses to activate ground floor uses.
And when somebody ridicules the idea of no ground floor use in a garage, ground floor business locates in the ground floor of garages all the time (even in Dallas - rarely successful because...) when demand driven by pedestrian activity allows it.