Whether partial updates are less subject to error or not depends on the design and implementation of the code. If object-oriented design and coding have been rigorously adhered to, along with an end to end test suite designed to break the code and not just prove that it works, a module by module update should be less likely to induce new errors and permit much faster turn around times.
That's a lot of "ifs" to get to a mere "should be."
And even iin the best of circumstances, why would that yield better results than the "clean installs" we're looking at?
