> Which lags behind upstream in versioning and has its own quirks.
No doubt re-implentation will lag and will not be perfect (either buggy, or not buggy enough to match reference). The big question: is it AWS users shopping for DBs, or MySQL/Postgres/CockroachDB users shopping for *aaS services? I suspect the former outnumber the latter, and they will likely tolerate being behind the bleeding edge.
I think it's still the majority of developers that first deal with databases outside a cloud environment and then want a managed version of it in their environment (which btw, Amazon does offer in RDS Postgres/MySQl, but it's their forks that have these problems).
And being behind the bleeding edge is problematic at least these days that databases are incorporating great features, but the bigger problem is incompatibility and subtle differences. Those are the truly annoying issues.
Which lags behind upstream in versioning and has its own quirks.
Or Redshift, which is still stuck in Postgres 8 and absolutely has its own quirks.