From chapter 3, I found this explanation of 'development' tracks of players rather thought-provoking:
Do online communities of other sorts have their own development tracks? StackOverflow, example, has a huge population of "browsers" and "students seeking homework help". Do they mature into other categories, or do those who are more active tend to ramp up right away? I know the site makes a lot of their info available so it might be interesting to datamine looking for a progression like the one Bartle describes for virtual worlds.
In the real world: does this have any relationship to why so few members of the U.S. Congress have science and engineering backgrounds? ;)
My other thought, just partway through the book, is his emphasis on planning and upfront design work, i.e., the waterfall. Is it possible to develop virtual worlds in a more agile way?
One obstacle is perhaps the large amount of content. Once you have a lot of areas, items, monsters, special abilities, etc. that are all working together based on a common set of assumptions, it doesn't really matter that you could refactor the code. How do you refactor the content? Obviously you could automate transformations of some of the content, but to players that might look like "you took my gear away!" Or "you broke all my scripts!"
The "minimum viable product" may also be quite large.