There's various kinds of documentation of differing granularity. Typically the high level documentation (coarse grained if you will) about architecture, requirements, etc. stays good over time but finer grain documentation (the kind I mentioned in my earlier comment -- data structures, order of function invocation, etc.) gets stale quickly because source code is constantly changing -- it gets refactored, plenty of bug fixes are put in, etc.

Reverse engineering is probably a loaded word, but in general digging in to understand the codebase helps you work more efficiently and avoid introducing bugs to the extent possible.