It looks like the SqlAlchemy uses their own native async engine.

        engine = sa_asyncio.create_async_engine(
            f"postgresql+asyncpg://sqlalch_bench:edgedbbenchmark@"
            f"{ctx.db_host}:{ctx.pg_port}/sqlalch_bench"
        )

If you subbed in your (ie MagicStack's) AsyncPG library instead instead (should be a couple line change), I'd be very curious to see a comparison... that would be a revealing benchmark to see how your EdgeDB Query Optimizer is doing (vs how much of the performance difference is better serialization).