I've had exactly the same situation, ~2M MAU service with REDIS as the only persistence system, all data being JSON serialized Pydantic models. The storage overhead was just terrible and cost real money.
This would have been a super nice to have back then.
I wonder though how much sense it would make to get something like this mainlined into upstream Pydantic? as having this downstream would give many continuity and dependency lock concerns. And having it as part of the main library would significantly drive adoption rate.
M68k has the advantage that it has a fairly typical memory model.
Alpha's memory model has problems with providing atomic access to single bytes, which i'd imagine in a kernel is a bit annoying :-)
And then there's just the social aspect, m68k was used in the Amiga/Atari/Mac/QL/x68k, so there is a whole generation of us m68k fans who are willing to keep it alive.
Alpha has it's fans (me included!), but it's not exactly the same. So in a way it's no surprise it's slowly bitrotting away.