The first cluster-aware version of RTNM automatically reloads all node types from the system projection whenever any node type is registered or unregistered elsewhere in the cluster. This is technically correct, but will typically require a lot of extra work. For example, even if only one node type is changed, all 40-odd built-in types as well as any custom types are reloaded. Since the RTNM write lock is held during this operation, this will greatly slow or stop other operations in the repository while the refresh is occurring.

It should be straightforward to only reload (or unregister) changed nodes instead. This will be a much more efficient implementation of this method.