IMO, it depends on the type of data as well. But given your read and write description basically any of the databases you listed is capable if configured properly. If you have a lot of experience in the SQL side, you may want to stick there to minimize time to develop your MVP.

I use MongoDB for our core database, but we also have data in elasticsearch for text based queries and some reporting. I am right now contemplating introducing Postgres for the accounting side of our SaaS and to possibly serve some of our reporting requirements that we have found are truly highly relational.

So I don't think in today's environment that you need to say one database is the end all. However, I agree that picking one is the right choice for a new project and then adding/changing as you find the requirements that don't fit. We are micro-serviced based too so in some ways it makes it easier because we isolate which services interact with which databases, but I won't say it isn't without its own complications too.