I don’t have a fundamental objection to using id for this, but I think you’re making this sound too simple:
Before turning an auto-generated, non-displayed variable into something that’s displayable and user-configurable (that can be duplicate in a document, e.g. if multiple people edit a document), it’d be worth checking with implementations on how they use it and if that’d cause problems.
I don’t think Zotero does much with the id – they use the uri they display outside of the schema JSON for linking, but Mendeley might, for example, and we don’t have much insight on other tools like paperpile and Readcube/Papers on this.
And even where the application doesn’t use it much, if this is going to be in 1.0.2 it may still require more changes to word processor integration to deal with old ids in documents, e.g., than would be desirable for a minor update.