It won't traverse across domains without an explicit argument, so it won't go too crazy. It also won't backtrack, so you won't end up with a complete copy of a blog if you've only bookmarked a single article. You can, of course, reduce the maxdepth to 1 pretty safely, if you don't think you'll need more than the specific bookmarked page.

I use that for programming language documentation - for example I'll hit the root page of Python's standard library documentation, and get everything I need locally.