You could always start with postgresql and use the transaction log to materialize views for particular use cases (ie differential data flow). This post is click

You could always start with postgresql and use the transaction log to materialize views for particular use cases (ie differential data flow). This post is clickbaity in that it doesn’t give any administration advice but right in the broad sense that postgres/mysql are still the best low-latency single sources of truth available