What’s the quickest way to create a Postgres Aurora instance with obfuscated production data?

I'm looking to speed up our process for creating a Postgres Aurora instance that uses obfuscated production data. Currently, we take full production snapshots, which include unnecessary data and empty space. The obfuscation job then restores these snapshots, scrubs sensitive information with SQL updates, and creates a new snapshot for use across dev and QA environments.

Since it's a monolithic database, I'm considering two ideas to enhance speed: using `pg_dump` instead of our full snapshot procedure or running a `VACUUM FULL` to shrink our obfuscation cluster storage prior to taking the final snapshot. For reference, our current stats show that a compressed `pg_dump` is around 15 GB, while RDS snapshots vary from 200 to 500 GB. Also, restoring a snapshot on Graviton RDS takes at least an hour, though it's quicker on Aurora Serverless v2.

So my question is: Is it a better route to pursue `pg_dump` for a faster restore, or should I focus on optimizing the obfuscation process and resizing the snapshot to a more manageable size, like 50 GB? And just to clarify, I'd prefer not to delve into splitting the database into microservices unless there's no other option.