The current definition is:
The name of the record used when updating or deleting a single record.

This implies it is only relevant for cross references inside one instance of a configuration file.

I am totally fine with the implication but other readers of the whole configuration concept may assume that the server has to preserve the names after the update is completed.

It may be impossible for a server to preserve the names if it has to fit the configuration into an existing configuration schema. A server should be able to convert from and to an existing configuration schema and should be able to create unique names in the configuration file but it may be not able to store the names provided by a configuration tool.

Since the names are used mainly for cross references and a client has to read the file after the update is completed to get a consistent version with the current configuration version, a client can adjust to names if they are changed by the server e.g. automatically created from an internal index.

For names that are relevant for configuration e.g. CertificateGroupDataType, we can explicitly state that the name must be preserved.