Uneven load distribution between threads on the TCP server

Hello. There was a problem with uneven load distribution on the TCP server. With the simultaneous mass addition of RTSP flows, the following problem arises

When adding a delay, the following distribution occurs

Is it possible to achieve such results without adding a delay?

I confirm. I observe the same uneven load when adding video streams very quickly

The load balancing algorithm of zltoolkit is not round robin,
So some imbalances are reasonable。

There is a thread load evaluation algorithm inside the zltoolkit, which has latency and may not be accurately evaluated in the short term.

This is the load evaluation algorithm:

ZLToolKit/src/Thread/TaskExecutor.h

Line 25 in e6fd7c7

class ThreadLoadCounter {

This is the algorithm for obtaining the thread with the lowest load:

ZLToolKit/src/Thread/TaskExecutor.cpp

Line 126 in e6fd7c7

TaskExecutor::Ptr TaskExecutorGetterImp::getExecutor() {

Thank you