Right, I resolved it at last:
Deploying a Zimlet in a "Cluster"( i.e. across replicas) is not as smooth as zmzimletctl indicates.
The script writes to stdout that the zimlet is being deployed to each individual server (according to it's package installation).
Us having 2 full installation 1 as LDAP master the other as LDAP replica, the script indicated:
Deploying Zimlet on server 1 and deploying Zimlet on Server 2. You are supposed to review mailbox.log on each server to check for success.
Logfiles on both servers did so.
As far as i can see, deploying through zmzimletctl only "activates" the Zimlet on these servers, but doesn't put the neccessary resources in place (i.e. files).
Hence, I undeployed the Zimbra Samba Zimlet on all servers which undid the file system changes. Then I restarted mailboxd on both machines to clear cached .
js (which seem to have caused the confusion). Then I deployed the zimlet using zmzimletctl on each server, ignoring the message that deploying it in one place would also deploy it on the other machines.
Once that ccompleted, everything worked nicely as expected.
One unsolved mystery remains though: Which setting tells ZimbraAdmin to load Zimlets-nodev_all.
js
This secret nedd to be revealed in another thread.
Even though I was the only contributing member to this thread, I hope this may be helpful for others (or myself in the future).