Bumped up minimum poll timeout in threadpool to address high CPU usage.

This should be exposed as a configuration option, since latency may become an issue in a client instances with 1000+ active drive connections.