There are tricks around known to solve this, such as building intermediate solutions like what Gates proposes to do.

When you have a serious issue in production that can be fixed with a global variable in 5 minutes or spending a day correctly planning and implementing a fix that routes the data correctly, what do you pick? I'd hope doing the first thing and replacing it with the second.

It's that sort of decision, only people die if you don't move in time.

Yes, the magic toilet took a possibly a decade to make, and I bet will never ever see real world deployment.

In the same time, most banal sewers and septic tanks take days to build.