In my current job we have lots of large and often 'sparsely populated' objects which means basically you can never count on any of the properties you were hoping would be present will actually be present. This means the data objects are essentially useless (at least for knowing what data you have to work with as you are writing code), and violate Pike's rules mentioned about data object design being critically important. In the modern world the younger generation of developers thinks "functional programming is great, and OOP (inheritance, etc) is obsolete", and only after 20yrs of coding they look back and eventually realize OOP had it right all along.
In architectures with massive numbers of sparsely populated objects only the guy who originally wrote any given function will be able to understand it, and everyone else who tries to work in the code is pretty much screwed.