Well, it's a live connection, so there could be a request at any time. While I don't know the internals and what the timeout might be for a live vs. extract data source, it doesn't sound that surprising that a frequently accessed near-real-time data source would have a persistent connection.
What's the reason for the concern over these connections? Most database platforms leverage connection pooling which keeps connections alive so they can be reused by other threads. Could that be what's happening here? Ultimately, I don't think this unclosed connection is anything to worry about.
Thanks Ken and Michael for your responses. Let me explain a bit further - I have a dashboard that ends up having 11 connections to the same db (don't get me started on why, bad data architecture / separate DBs, going away soon) and am creating views such that I have 10 filtered views being sent via email subscriptions. As I run these subscriptions, each creates 10 separate connections, creating about 110 connections as they run. Persisting these connections would be fine if each view reused the original 11 connections; but each view is creating its own 11 and exhausting the connection pool.
I see that the timeout for a live connection is about 20 mins with MySQL, so I can probably architect around it. My surprise is mostly regarding different views of the same dashboard not reusing existing connections.