GitLab

Backends should just execute when invoked, frontend should
exclusively take care of queueing for txn setup.

Since c_mutex is now always inited at startup time, we no longer
need connections_mutex to protect from uninit'd structures

Schedule a new retry instead of immediate looping back. This
will also prevent locking up cn=config replication.

See if they were requested as soon as they are successfully registered.

Do it explicitly in servers

Also make sure config_add_internal errmsg is returned to slapadd

send_ldap_response() clears them immediately even if we never attached
the data to be freed, so when we reinstate them, the flags are gone and
the next send_ldap_response() doesn't consider freeing them.