Webkit started as a fork of KHTML (part of KDE and LGPL licenced) their mishandling of it is a good example of how not to work with open source.
Andriod is another. As Google isn't developing it in public they're carrying the full cost. And their re-creating of userland libraries to avoid the LGPL is pretty much the opposite of your argument (though possibly inherited from before Google bought Android).
Aside from your examples I agree with your point. It's similar to one Doctorow makes in Makers that everything turns into a commodity after a few years and it doesn't make you money anymore. However it can still act as a barrier of entry so examples like Chadaustin's can still make sense if maintenance costs don't dominate.
Ironically Chrome OS is Ubuntu (and Chrome is based on Webkit) so they are also examples where commodification has already happened.
I've seen this claim quite a few times ("the service model only works for Red Hat") - is it actually true though?
Sure, there aren't that many billion dollar companies built around services, but I've seen dozens of smaller lifestyle businesses built around consulting against open source projects.
Off the top of my head...
Hwaci for SQLite: http://www.sqlite.org/support.html
Dozens of companies for PostgreSQL: http://www.postgresql.org/support/professional_support
Revolution Systems for Django: http://www.revsys.com/services/django/
Spring Source for Spring: http://www.springsource.com/
Lucid Imagination for Solr/Lucene: http://www.lucidimagination.com/
Lemur Consulting for Xapian: http://www.lemurconsulting.com/
Then there's the fact that a bunch of the current NoSQL efforts (MongoDB for example) are getting funding, which suggests investors still believe you can build a business around open source support.
At the local, per-company level, these anti-open-source policies may have made sense. However, they had some global effects:
* the industry wasted tons of money reinventing solutions to solved problems.
* to start a AAA game studio, you must license or invent graphics, physics, threading, etc., too expensive for long-tail innovation.
By leveraging and encouraging open source, you can lift all boats, improving the world in which we compete.
In short, on my deathbed, leveraged open source means a lot more than closed source would.
Early Id, Unreal & Duke Nukem 3d are all OSS now.
By the way, the other engines you mentioned are all geared towards first-person shooters.