How to definitively identify company officers across snapshot (Prod195) and streaming api data?

Not from Companies House and I can’t provide anything helpful - just to echo your issue is known and not new. Indeed, there have been such notes and queries pretty much since the inception of this API. From 2019 for example:

As far as I’m aware similar issues apply both within some other datasets (e.g. certainly PSCs) and between all the “people” datasets - Officers (presumably “Officer appointments” - which a company Officers record just references - hence multiple “the same person” entries within a Company?), PSCs and Disqualified Officers.

Good luck.