File Metadata API: Adding or updating file metadata doesn't create a dataset draft, rather it fails with filename exists.

If I try setting the file metadata of a particular file on a published dataset, it fails with

curl -H "X-Dataverse-key:XXX" -X POST -F 'jsonData={"description":"My description bbb.","provFreeform":"Test prov freeform","categories":["Data"],"dataFileTags":["Survey"],"restrict":false}' "http://localhost:8080/api/files/:persistentId/metadata?persistentId=doi:10.70122/FK2/HF27IA/FHGZJM"

{"status":"ERROR","message":"Filename already exists at 50by1000.tab"}

But if I first create a draft of the published datasets and rerun the command, it works. It's my expectation based on experience and UI behavior that updating any metadata should create a new dataset draft if one doesn't yet exist.

Start with the File Metadata API endpoint shown in the curl example and reproduce the request against a published dataset without an existing draft. Trace how metadata updates handle the dataset version and filename conflict; done means the update creates a draft instead of returning the filename-exists error.