Make Data Nodes aware of distribution function (such as hash) so it can check if the values being used violate the rule or not.
This could happen with a misconfigured postgresql.conf on a coordinator, or for rules and, in the future, trigger functions.
It is also related to handle distribution column value update and tuple relocation. If datanode is aware of distribution function, it can ship tuples to other data node, which can be done directly (peer to peer) or through coordinator.
This happens only when applications connect to datanodes directory. In this case, warning will be shown and it is highly discouraged to do so, especially issuing DDL amd DML.