Metrics are generated each scrape, causing significant churn. If metrics could be initialized when loaded from a file, this could significantly simplify the scrape. A per-database scrape would then only query the database for the metric, update the metric value, and send it to the collector.