This is probably the more important thing about Linuxconf and the most difficult thing to achieve for an administration framework. This explain why no framework (on any OS ?) is scratching the surface of this. What does it means ?
To achieve this, all components of Linuxconf, including modules, were asked one by one to check various things. For many modules, this is very simple: They check the dates of various config file and trigger a restart on the appropriate daemon if its starting date is older than the configuration file. In other case, this means probing various kernel structure (in /proc) and updating them with care.
And all this is done in either simulation mode, where the component shows what it will do and in real mode, where it does it. This processing take about a second on a typical workstation or server.
Although unrelated in term of facility in Linuxconf, both technology rely on the ability to activate configuration changes on the fly. Without that, most users will end up rebooting "to make sure". If you are remote managing a large amount of stations and servers, rebooting them is often not possible.