[PaulMest]

I'd guess that https://neon.tech/ could help with this. Neon is branded as "Serverless Postgres". They have APIs to create branches of your database.

So you could effectively:

1) Create a snapshot of your production DB -> DB_2A

2) Then create a snapshot of that snapshot -> DB_2B

3) Now you have two copies of the exact same database. Run your query/workload/migration on DB_2B.

4) Run some metadata queries against DB_2A and DB_2B and compare the results.

5) If your metadata queries are inline with expectations, delete the snapshots. If not, leave them around for a bit for manual inspection.

[pawelduda]

Interesting. Will check it out in detail. From a quick glance tho I'm afraid that vendor-specific cloud is a no-go because I would rather not upload any client's data in there, even if anonymized. Besides, it's an overkill to convince anyone to switch to a different vendor for sake of diffing DB alone :)