The argument for choosing dull tools is not that new tools are bad. It is that novelty has a carrying cost paid in operational attention, and that budget is small.

Counting what it costs

A new datastore is not one decision. It is a backup story, a monitoring story, an on-call runbook, a scaling model, a failure mode nobody on the team has seen yet, and an upgrade path. PostgreSQL already has all of those written down and half your team has been paged by it before.

Spend the novelty budget where it differentiates the product. If the interesting thing about your system is the domain model, then the queue, the database and the deployment pipeline should be aggressively unremarkable.