We run Postgres/Drizzle in production and tuning params (shared_buffers, work_mem, etc.) has always been trial-and-error for us. Curious what workload characteristics it uses to auto-tune — read/write ratio, connection count, something else? Also curious how it compares to just running pgtune once and calling it done.