Actually that’s quite opposite to philosophy of having thread pool, idea with thread pool is to have minimum no. of threads running for serving ‘n’ no. of connections. Basically all the connections are spread across thread groups and scheduler initializes each thread group with single thread called ‘listener’ thread, it listens for incoming statement request and tries to execute it immediately if it can else it puts it in a queue to be executed later. Statement waits in queue for specific time and then a new thread is spawned to execute the statement.

In other words, with thread pool enabled, max_threads = max_connections + no. of thread groups (in worst case scenario)