This resonated strongly with me. Almost all of the heavy DB interaction projects I've done in the last couple of years - be it in Python, Go or Java, the SQL layer has almost always been raw queries going to the DB. Makes things so much easier.
I don't miss SQLAlchemy where I would trawl through documentation to figure out how to do X or Y.
As for SQL standardization, +1 to that. Learning a new pseudo-SQL is only fine if there are additional features on top of the DB, but I've encountered databases where the SQL support was present in the name only and extremely limited, with people making do with UDFs in order to achieve basic things.