Of course, people do defend dissertations based on work that was a collaboration with others, but I think we're considering a level of collaboration not usually seen.
Is the effort expended for the paper able to be divided cleanly into chunks 1/20th the size, to be credited to each author in turn?
Doing research, or any creative work, is not a manufacturing process; the output does not inevitably scale linearly based on the number of inputs. It is possible that a paper could describe something that was very impressive for 20 people to accomplish.
In general it only counts on your CV if you're the first, second, or maybe third author. But a project with 20 people on it would probably collapse under the organizational weight before completion if it wasn't a pretty big deal. For example, the papers on the Human Genome Project and shotgun sequencing had a lot of authors, with the breakdown as described above.
That still leaves the problem of reputation-building; a paper where an individual is only one of 20 authors probably isn't so impressive on a CV.