The idea you are proposing, is a nice one. However we don't think we
should implement this in the treeview, because this will cause lots of
unwanted behaviour. For example, a filesystem model only reads the
next directory if the node for that directory is being expanded. So if
you're going to search through all unexpanded trees, it may start
reading every directory on your hard disk, on big volumes this will
take ages. Other than that it will use a lot of memory.

So this is not something we want to happen. The same reason is valid
for the gconf-editor tree (reading the complete gconf-editor tree
isn't really fast either), and most other models where this feature
could be useful. Not all of them though.

In summary, I think it's a nice idea, but it's not a good idea to
implement it as it will probably cause a lot of problems.