I have the impression PARTY class got a new 'redefine' of uid attribute somewhere between v1.0.0 and v1.0.1. Previously was addressed or touched in SPEC-194.
The only thing that mentions something about uid (indirectly) is the following text from section https://www.openehr.org/releases/RM/Release-1.0.2/docs/demographic/demographic.html#_party_identification:
"Identifiers of Party instances in the system are provided in the same way as identifiers of Compositions in the EHR: via the uid attribute (type OBJECT_VERSION_ID) of the containing VERSION. "
If I'm not mistaken this is just an export problem of UML, than propagated to specs.
I think the correct type is the one that inherits from LOCATABLE (so UID_BASED_ID). I'm not sure though if this attribute should be required or not, overriding optional state from LOCATABLE. At least Code24 always implemented this as specified in from LOCATABLE.