@lucas.lima
I think that’s something specific to AWS RDS. If you have more control over your database, then you can tell it to kill connections after X minutes. Metabase only handles 15 connections per database (a database counts as defined in Admin > Databases).
I don’t have long-running queries, so I have never really encountered this problem. When I have queries that takes more than 30 seconds, then I save the output to a view, so the end-user gets a response instantly from the view, but the data may be a few minutes old.
EDIT: The “safe” way to test this, would be to change the limit to let’s say 20 and just be prepared to go back to default, if you’re seeing any strange behavior.