Refactor the update script so that there are two scripts - one internal one that does the update for one layer at a time, and another that does the git fetching and runs the internal script for each layer. This will eliminate the memory leaks and other issues we've had as the index grows once and for all, as well as making it so we don't have to edit the script for every YP release.
This has now been implemented. Adding a new release is now as simple as adding a branch in the admin interface - no need to modify anything on the backend.