This is, in my opinion, the primary strength of tags over folders. It’s easier to modify the tags of a document than to deal with a scattered few of replicants. Merging tags is straightforward and effective maintenance. The fact that we have a tendency to over-tag things illustrates the flexibility of tags.
Tags are also more search-friendly. It is trivial to search for all items tagged apple, banana and candy. Finding all items with a replicant in each of three folders, however, can be a bit more tricky.
A stark contrast is in the decimal system used by @Yclipse , which is folder-based, single-threaded, and optimized for manual retrieval of physical items. Decimal systems are fundamentally based on the premise that you can determine whether something (books, in the case of a DDC library) suits your needs by its name/appearance alone. They do not take advantage of modern computers’ capabilities to process and combine multiple threads of information.