Just a note that the University of Colorado at Boulder is abbreviated as CU not UC.

The XML databases and tools are just now becoming mature enough to compete on the same level as relational databases. I would say they won't completely replace them but they have the potential to augment them in some interesting verticals. Anywhere the problem can be thought of as document centric (or hierarchical) an XML DB has the potential to out preform a relational data-store.

For some examples of XML DBs you can refer to:
http://exist.sourceforge.net/
http://www.modis.ispras.ru/sedna/
http://www.marklogic.com/
http://www.x-hive.com/

Also take a look at http://www.saxonica.com/ (Michael Kay's company) for some insight on how XML DB's can be used.
Remember that it takes years for major process cycles to change. XML DB's may not be in the limelight yet but their time is approaching.

At 200-500 pages docbook would work well for a single purpose document that doesn't need a lot of maintenance over time and is for a single purpose (e.g. never going to use sections for a overview or presentation). I still maintain however that the flexibility of DITA is better for any writing that my have more than one use. DITA however is far from perfect as a spec and I could list a number of things that could have been done better (same as with docbook). DITA also gives the ability to extend it's spec (schema's/dtd's) for your own use in a concise manner that can help with the type of task you mention. I'm not on the DITA project so take everything I say with a grain of salt.

The software my company develops can handle most specs so we've dealt with both and I just know that one monolithic schema/DTD presents its own unique problems over a distributed one. The same type of problems exist with monolithic vs. distributed documents (e.g. a single chapter document vs. a chapter as a collection of concepts). It's up for the project to decide which gives more benefit; However for a team larger than a handful, or when maintainability is important I think distributed gives the greater benefit.

Ok enough of that from me.

I understand your reservations about XMetal, I'm a software guy and prefer oXygen too, however a lot of our customers say that it is still too geared towards power users. In other words if your comfortable in a programmers IDE (e.g. eclipse) you'll be fine with oXygen but the tech writers that came from paper or MS Word seem to have lots of trouble with a tool like oXygen (and yes I mean the author edition).

I deal with airline and oem manuals for my job and docbook breaks down for those mostly because of the amount for process involved and the need for reuse. We use DITA, S100D and the ATA spec for these giant manuals (e.g. over 100 megs of text per manual). The more authors with diverse sets of data the better a system like DITA or S100D is (allows for breakup of the data in a logical way). This type of authoring allows for lots of people to write consistently (still needs a good style guide but it helps).

Oh and there are other, more author friendly, editors if you need those too for DITA like the XMetal Author (DITA Edition) from Just Systems.

While docbook can be good for a small setup it tends to be overly simple when working with large documents. I'd suggest instead using DITA as it allows for the types of referencing mentioned and also allows for pretty extensive reuse of content. Like docbook it can be transformed into most formats you may need but gives the added bonus of being able to break it up and re-organize the book structure dynamically (this is really how technical writing should be).

A good resource for how to use this powerful language can be found at http://dita.xml.org/. As the parent suggested <oXygen/> for XML is a great editor for this kind of work and comes with a DITA edition.

Also as the parent suggested svn is really the best way to go for revision control but a database like eXist can be a great resource management tool for your content while your working with it.

A few years back a Boulder, CO based company by the name of Network Photonics had successfully created an OOO switch however the telco situation at the time forced them to shut their doors. However it was very nice piece of tech that implements on method of not having to do an electric step in switching (not quite the same method as described in the article).