seamless_database_pool | Add support for master/slave database clusters | Database library

I have a Rails/ActiveRecord application that occasionally executes some heavy read-only queries against the application's models for ad-hoc admin/reporting purposes. I currently send these queries to a separate read-replica database endpoint using the seamless_database_pool gem so these queries do not add load to the primary database. However, this setup adds a constant runtime dependency on the replica, reducing my application's availability - whenever the replica goes offline for maintenance, the entire site returns 503 errors on any request, even though most requests never query the read-replica at all.

How can I connect to a read replica for ad-hoc reporting queries on ActiveRecord models without affecting the availability of my application? Is there another database-adapter gem that handles this use-case explicitly?

Or, (even better) is it possible to accomplish this directly using ActiveRecord? I know that it's possible to connect to a separate database by calling #establish_connection in order to direct a specific model to a separate database across all controller actions, but for my use case I need to connect to a separate database across all models, but only for specific controller actions.