I think his point is precisely that VCSs cannot automatically avoid semantic conflict. He's not blaming the VCS, but rather saying that Feature Branch should be avoided even if you're using a modern VCS.
This only becomes apparent in the footnotes where he states that if your feature branches live for only a day, then your basically doing what he calls continuous integration.
This feels more like something that can be remedied through having good practices. Just like you shouldn't go without committing for days when working with Subversion, it is probably a good idea in a DVCS to:
- integrate the main branch with your feature branch often (at least daily)
- don't do semantic changes in a feature branch. Instead, switch to the main branch (or a very shortlived new branch) and do the semantic change.
This should avoid 2 feature branches doing semantic changes and then having a merging hell after a week or so of development.