At one time the problem was that he wanted to tour with the machine, so it had to withstand transportation.
But in some recent videos he mentions how the form factor and the goal of "looking cool" was more important than reliability.
IMO the problem he is not really picking a consistent direction. Either make it road-worthy or make it look like a steampunk Rube Goldberg contraption. Or better: make two machines. But making a single machine that is both reliable and Rube-Goldberg-esque is be 10x more difficult than making two machines.
There's always the third option: first make a simple machine that works but looks crap, and then gradually upgrade the aesthetics of the worst looking part until it's "cool enough".
I explained it elsewhere, but the "looking cool" part is not just about aesthetics in the superficial sense. It's about the fact he wants to use marbles to trigger electro-acoustic instruments, and he wants to have some features that kinda suck, such as the "marble elevator".
He could keep the mechanical sequencer and trigger instruments using hammers. That would make this project more feasible. But then he wouldn't have the marbles...
Honestly all the problems here are centred around the "marbles" constraint :/
But your point stands: he could lower the scope and actually finish something.
So for the MM3, he decided to do more engineering and less good content. His videos are less interesting since, and contain less music.