If a developer uses the eZ cluser API to store any data in var/storage subdirectories, everything works fine up to the point where the clusterize script is used.
If the script is used to declusterize then reclusterize, all such data will be gone.
This is because the clusterize script only checks for binary content from existing attributes (ezimage, ezmedia, ezbinaryfile).

It would be nice to
1. document this
2. allow a switch in the operation of the clusterizing script which will make it crawl the whole db (going out) or fs (going in) to make it cope with data added by the developer