What determines the default timeout for keep-alive connections in Nimble?

Nimble utilizes a non-blocking I/O architecture to manage high-concurrency HTTP services. While the framework supports HTTP/1.1 keep-alive headers to minimize handshake overhead, the specific mechanism for reclaiming idle sockets is not explicitly defined in the high-level API.

In environments with high connection churn, managing the lifecycle of these persistent sockets is critical to prevent resource exhaustion. There is uncertainty regarding whether the server relies on a global timeout setting or if the lifecycle is governed strictly by the client's Keep-Alive header values.

Which configuration parameter controls the server-side idle timeout for persistent connections? Does Nimble provide a way to programmatically limit the maximum number of concurrent keep-alive sockets to prevent file descriptor exhaustion?