Use cache of cloned repos instead of cloning them multiple times
In the above example, extracted from https://forge.extranet.logilab.fr/cubicweb/cube-doctor/-/jobs/419718/ , it seems that addressbook cube is cloned 9 times in different /tmp folders.
Could there be an optimisation where the first clone is in /tmp/abcd, then the other clones do so from /tmp/abcd instead of from heptapod ?