Similar to Managing timeouts we might want the ability to customize the resources for given actions to prevent critical path code from failing.

• Limit the number of connections per endpoint. If you have some long-running operations consider queuing up connections instead of flooding the server with long running operations. If the connection pool is shared with higher throughput and lower latency endpoints you could block critical path code if too many long-running connections are being processed.
• Multiple connection pools (HTTP, JDBC) same as above but for other types of resources.