Sure, why not? Never reaching 100% compatible is a problem as is never amounting to more than a bug for bug clone. If they can quantify exactly what they have changed for good reason then that's good for them.
Sure, I'm impressed that something like this that would have been a thought experiment or a decades long project with minimal coverage is now a prototype that seems to me like it will contribute something back even if there might not be a rush to use it directly. During the NoSQL phase one can say it was real human engineers making that stuff but often the overall quality was bellow anything that should pass a quantitative measure like a postgres test suite and people put those in production even when they were known to have deep design flaws.
So even if they reach 100% compatibility they never get equality.
Then Vibe into that and add changes that breaks the existing coverage...
Why not reimplement the entire thing from the ground up with the changes in mind from the start instead?
The upgrade is an afterthought to a system that is not verified to be fully working and the new features will break the existing verification too.
If I had OCD it would be triggered to the max by this. But I'm just sipping a drink from the sidelines.