It's definitely really important that a portable database format that can be persisted on disk be used for the wallet file database, or else you couldn't move wallet files around and they'd (for the most part) be stuck on the system they were created in.
Actually, I think requiring a wallet file that can be moved around in all cases might actually hinder proper research on database backend efficiency. Ideally, database engines would be pluggable in the code, and there would be a database migration tool to switch between engines. Then it suffices to have one engine around that has a fixed wallet file format.
This would enable benchmarks like
https://blog.lopp.net/2020-bitcoin-node-performance-tests/ to be ran on one engine, but with different database backends enabled.
Being able to make informed decisions based on benchmarks will gain importance with the growth of the blockchain.