[metadat]

You can use a DB connection pool for this.

But how much request time will it really save? Have you measured the impact of that stage of the pipeline?