Rudder

When we copy ncf on the nodes, we use a very naive method of checking every files, which uses network ressources and CPU

The solution to be used is to make at every run a hash of ncf, and compare the hash of ncf with what we have on the node. If they differ, then we copy the promise
The risk there is that promise were regenerated, but hash not yet computed, so in this case, we'll copy ncf if promises has been updated

Note: the relay will always check for update of ncf, to prevent delays that may arise between hash update and real update