139行加了个sleep 10，配合229的sleep10 和 64行的KEEP_ALIVE_TIME=6 ，可以“模拟”特殊调度顺序。预期结果：第二次加的任务f2，一直不会执行

shrinkable_threadpool_test_1

"""
史上最强的python线程池。

最智能的可自动实时调节线程数量的线程池。此线程池和官方concurrent.futures的线程池 是鸭子类关系，所以可以一键替换类名 或者 import as来替换类名。
对比官方线程池，有4个创新功能或改进。

1、主要是不仅能扩大，还可自动缩小(官方内置的ThreadpoolExecutor不具备此功能，此概念是什么意思和目的，可以百度java ThreadpoolExecutor的KeepAliveTime参数的介绍)，

2、非常节制的开启多线程，例如实例化一个最大100线程数目的pool，每隔2秒submit一个函数任务，而函数每次只需要1秒就能完成，实际上只需要调节增加到1个线程就可以，不需要慢慢增加到100个线程
官方的线程池不够智能，会一直增加到最大线程数目，此线程池则不会。

3、线程池任务的queue队列，修改为有界队列

4、此线程池运行函数出错时候，直接显示线程错误，官方的线程池则不会显示错误，例如函数中写1/0,任然不现实错误。
"""
import os
import atexit
import queue
import sys
import threading
import time
import weakref

from nb_log import LoggerMixin, nb_print, LoggerLevelSetterMixin, LogManager


_shutdown = False
_threads_queues = weakref.WeakKeyDictionary()


def _python_exit():
    global _shutdown
    _shutdown = True
    items = list(_threads_queues.items())
    for t, q in items:
        q.put(None)
    for t, q in items:
        t.join()


atexit.register(_python_exit)


class _WorkItem(LoggerMixin):
    def __init__(self, fn, args, kwargs):
        self.fn = fn
        self.args = args
        self.kwargs = kwargs

    def run(self):
        # noinspection PyBroadException
        try:
            self.fn(*self.args, **self.kwargs)
        except BaseException as exc:
            self.logger.exception(f'函数 {self.fn.__name__} 中发生错误，错误原因是 {type(exc)} {exc} ')

    def __str__(self):
        return f'{(self.fn.__name__, self.args, self.kwargs)}'


class ThreadPoolExecutorShrinkAble(LoggerMixin, LoggerLevelSetterMixin):
    # 为了和官方自带的THredpoolexecutor保持完全一致的鸭子类，参数设置成死的，不然用户传参了。
    MIN_WORKERS = 5
    KEEP_ALIVE_TIME = 6

    def __init__(self, max_workers=None, thread_name_prefix=''):
        """
        最好需要兼容官方concurren.futures.ThreadPoolExecutor 和改版的BoundedThreadPoolExecutor，入参名字和个数保持了一致。
        :param max_workers:
        :param thread_name_prefix:
        """
        self._max_workers = max_workers or 4
        self._thread_name_prefix = thread_name_prefix
        self.work_queue = queue.Queue(max_workers)
        # self._threads = set()
        self._threads = weakref.WeakSet()
        self._lock_compute_threads_free_count = threading.Lock()
        self.threads_free_count = 0
        self._shutdown = False
        self._shutdown_lock = threading.Lock()

    def _change_threads_free_count(self, change_num):
        with self._lock_compute_threads_free_count:
            self.threads_free_count += change_num

    def submit(self, func, *args, **kwargs):
        with self._shutdown_lock:
            if self._shutdown:
                raise RuntimeError('不能添加新的任务到线程池')
        self._adjust_thread_count()

        self.work_queue.put(_WorkItem(func, args, kwargs))

    def _adjust_thread_count(self):
        # if len(self._threads) < self._threads_num:
        # self.logger.debug(
        #     (self.threads_free_count, len(self._threads), len(_threads_queues), get_current_threads_num()))

        if self.threads_free_count < self.MIN_WORKERS and len(self._threads) < self._max_workers:
            # t = threading.Thread(target=_work,
            #                      args=(self._work_queue,self))

            t = _CustomThread(self).set_log_level(self.logger.level)
            t.setDaemon(True)
            t.start()
            self._threads.add(t)
            _threads_queues[t] = self.work_queue

    def shutdown(self, wait=True):
        with self._shutdown_lock:
            self._shutdown = True
            self.work_queue.put(None)
        if wait:
            for t in self._threads:
                t.join()

    def __enter__(self):
        return self

    def __exit__(self, exc_type, exc_val, exc_tb):
        self.shutdown(wait=True)
        return False


# 两个名字都可以，兼容以前的老名字（中文意思是 自定义线程池），但新名字更能表达意思（可缩小线程池）。
CustomThreadpoolExecutor = ThreadPoolExecutorShrinkAble


# noinspection PyProtectedMember
class _CustomThread(threading.Thread, LoggerMixin, LoggerLevelSetterMixin):
    def __init__(self, executorx: ThreadPoolExecutorShrinkAble):
        super().__init__()
        self._executorx = executorx

    def _remove_thread(self, stop_resson=''):
        # noinspection PyUnresolvedReferences
        self.logger.info(f'停止线程 {self._ident}, 触发条件是 {stop_resson} ')
        self.logger.info(f'sleep模拟调度')
        time.sleep(10)
        self._executorx._change_threads_free_count(-1)
        self._executorx._threads.remove(self)
        _threads_queues.pop(self)

    # noinspection PyProtectedMember
    def run(self):
        # noinspection PyUnresolvedReferences
        self.logger.debug(f'新启动线程 {self._ident} ')
        self._executorx._change_threads_free_count(1)
        while True:
            try:
                work_item = self._executorx.work_queue.get(block=True, timeout=self._executorx.KEEP_ALIVE_TIME)
            except queue.Empty:
                # continue
                # self._remove_thread()
                # break
                if self._executorx.threads_free_count > self._executorx.MIN_WORKERS:
                    self._remove_thread(
                        f'当前线程超过 {self._executorx.KEEP_ALIVE_TIME} 秒没有任务，线程池中不在工作状态中的线程数量是 '
                        f'{self._executorx.threads_free_count}，超过了指定的数量 {self._executorx.MIN_WORKERS}')
                    break  # 退出while 1，即是结束。这里才是决定线程结束销毁，_remove_thread只是个名字而已，不是由那个来销毁线程。
                else:
                    continue

            # nb_print(work_item)
            if work_item is not None:
                self._executorx._change_threads_free_count(-1)
                work_item.run()
                del work_item
                self._executorx._change_threads_free_count(1)
                continue
            if _shutdown or self._executorx._shutdown:
                self._executorx.work_queue.put(None)
                break


process_name_set = set()
logger_show_current_threads_num = LogManager('show_current_threads_num').get_logger_and_add_handlers(
    formatter_template=5, log_filename='show_current_threads_num.log', do_not_use_color_handler=False)


def show_current_threads_num(sleep_time=600, process_name='', block=False, daemon=True, work_queue=None):
    process_name = sys.argv[0] if process_name == '' else process_name

    def _show_current_threads_num():
        while True:
            # logger_show_current_threads_num.info(f'{process_name} 进程 的 并发数量是 -->  {threading.active_count()}')
            # nb_print(f'  {process_name} {os.getpid()} 进程 的 线程数量是 -->  {threading.active_count()}')
            logger_show_current_threads_num.info(
                f'  {process_name} {os.getpid()} 进程 的 线程数量是 -->  {threading.active_count()}, work_queue剩余任务数->{work_queue.qsize()}')
            time.sleep(sleep_time)

    if process_name not in process_name_set:
        if block:
            _show_current_threads_num()
        else:
            t = threading.Thread(target=_show_current_threads_num, daemon=daemon)
            t.start()
        process_name_set.add(process_name)


def get_current_threads_num():
    return threading.active_count()


if __name__ == '__main__':
    from concurrent.futures import ThreadPoolExecutor

    pool = ThreadPoolExecutorShrinkAble(10)
    # pool = ThreadPoolExecutor(200)  # 测试对比官方自带

    show_current_threads_num(sleep_time=5, work_queue=pool.work_queue)


    def f1(a):
        time.sleep(0.2)  # 可修改这个数字测试多线程数量调节功能。
        nb_print(f'{a} 。。。。。。。')
        # raise Exception('抛个错误测试')  # 官方的不会显示函数出错你，你还以为你写的代码没毛病呢。

    def f2(a):
        time.sleep(0.2)  # 可修改这个数字测试多线程数量调节功能。
        1/0


    for i in range(30):
        time.sleep(0.01)  # 这里的间隔时间模拟，当任务来临不密集，只需要少量线程就能搞定f1了，因为f1的消耗时间短，
        # 不需要开那么多线程，CustomThreadPoolExecutor比ThreadPoolExecutor 优势之一。
        pool.submit(f1, str(i))
    #
    time.sleep(10)
    for i in range(30):
        pool.submit(f2, str(i))

    time.sleep(100000)

所以不是纯粹的调用方式复现的问题，而是修改了作者的源码（即139行的sleep）来放大问题。仅做研究之用吧。

加q 909686719