I think that PostGIS in Action filled a big hole and I find it very useful as each chapter comes out, http://www.postgis.us/

Books I'd like to buy that don't exist:
OpenLayers cookbook, specifically combining all the existing individual examples into a comprehensive application. Each example is easy to do by itself, combining them gets more difficult (for me). Perhaps this book could include GeoExt and other JS libraries commonly used with OpenLayers. Some templates and best practices would be very helpful.

The stack in production:
There are endless things that I have set up in development and many of these are getting easier to set up. What are the steps that I need to take to get this to production level? If I am going to deploy Postgresql/Postgis, Mapserver or Geoserver, and OpenLayers/GeoExt how do I spec hardware, set configurations files for memory allotment and other variables, ensure security, etc? Some of this is very particular to each situation, however, much of it may be similar. There could also be ranges on the number that corresponds to ranges of number of hits/month, concurrent users, etc.

I prefer these books to be accessible from a GIS background, since that is my situation and I think many others as well.

What about writing custom geoprocessing scripts in python with shapely, gdal, numpy, rpy, ...

With complex, usable vector operations like tracing a line from gps-tracking, with a new feature every km, adding an average time, joining events as attributes that have happened in that period, ...
And with raster operations like classifying, clustering, smoothing, feature extraction ...

Setting it up as a WPS, integrating it in a Openlayers-webApp, in qgis ...
Maybe also something about jython-geoserver.

There are no real books about this, no with python anyway, while python is becoming so popular for this kind of thing.