Personally, I am firmly in the second camp (reading multiple config files). My reasoning is as follows.

I am working on a large project that will result in a common framework, and product modules which plug in to that framework. The common framework, and all of the product modules, will use Struts. Product modules can be installed and uninstalled by our customers.

Using some kind of 'include' mechanism solves the problem of different product teams stepping on each others' toes when editing the Struts configuration. However, it makes installation and uninstallation somewhat harder, because those applications will have to understand the structure of the Struts configuration file.

Using multiple configuration files allows me to have one config file for the common framework, and then one more for each product. Each product team can concentrate on its own config file, and the installer and uninstaller only have to worry about which files to add or remove, and not about what they contain or how to modify them.