Each Darcs repository contains an inventories subdirectory. Each file inside this directory is an inventory, that is a gzipped list of patches. These files accumulate over time. For instance, on a repository with 193 patches and no tag, I have 101 inventory files, totalizing 1 MByte. Each one of these files is the previous one plus the description of one patch. Now, if I do a “darcs get” of this repo, the new repo has no inventory at all. Indeed, deleting all these files gives me a fully working darcs repository.

This is because for each patch added to the repository, a new inventory file, representing the state of the repo, is created and added to _darcs/inventories/. Tagging the repository does not make these inventory files go away.

The last two paragraphs of darcs help optimize clearly tell that doing a darcs get is the only solution to solve this problem.