Williams was worried about the current draw on the switch matrix which is why they had resistors on the early driver boards in this circuit - and probably also why they did the momentary drop switch with the all-done signal like an EM would. The advantage to this setup is that the all-down signal does not depend on the game 'seeing' all the momentary switches. Bally/stern games if they miss a drop or the switch is dirty/not registering the bank could go down and stay down. (Later sterns mpu200 games solve this a little by letting drop switches reactivate, but since they were 'marked' closed, they do not score again.)
I don't see why the switch matrix couldn't read every single switch down physically. Sounds like an interesting experiment to take some jumper wires and connectors and try it out. I do notice that later games sometimes 'stagger' the switches that are likely to be closed so that they aren't all in the same column/row. Likely they were being overly cautious about how much current the transistors and chips could sink.