How to terminate a Uvicorn + FastAPI application cleanly with workers >= 2 when testing with pytest

I have an application written with Uvicorn + FastAPI.
I am testing the response time using PyTest.

Referring to How to start a Uvicorn + FastAPI in background when testing with PyTest, I wrote the test.
However, I found the application process alive after completing the test when workers >= 2.

I want to terminate the application process cleanly at the end of the test.

Do you have any idea?

The details are as follows.

Environment

• Windows 10
• Bash 4.4.23 (https://cmder.net/)
• python 3.7.5

Libraries

• fastapi == 0.68.0
• uvicorn == 0.14.0
• requests == 2.26.0
• pytest == 6.2.4

Sample Codes

• Application: main.py

  from fastapi import FastAPI
  
  app = FastAPI()
  
  @app.get("/")
  def hello_world():
      return "hello world"
  

• Test: test_main.py

  from multiprocessing import Process
  import pytest
  import requests
  import time
  import uvicorn
  
  HOST = "127.0.0.1"
  PORT = 8765
  WORKERS = 1
  
  
  def run_server(host: str, port: int, workers: int, wait: int = 15) -> Process:
      proc = Process(
          target=uvicorn.run,
          args=("main:app",),
          kwargs={
              "host": host,
              "port": port,
              "workers": workers,
          },
      )
      proc.start()
      time.sleep(wait)
      assert proc.is_alive()
      return proc
  
  
  def shutdown_server(proc: Process):
      proc.terminate()
      for _ in range(5):
          if proc.is_alive():
              time.sleep(5)
          else:
              return
      else:
          raise Exception("Process still alive")
  
  
  def check_response(host: str, port: int):
      assert requests.get(f"http://{host}:{port}").text == '"hello world"'
  
  
  def check_response_time(host: str, port: int, tol: float = 1e-2):
      s = time.time()
      requests.get(f"http://{host}:{port}")
      e = time.time()
      assert e-s < tol
  
  
  @pytest.fixture(scope="session")
  def server():
      proc = run_server(HOST, PORT, WORKERS)
      try:
          yield
      finally:
          shutdown_server(proc)
  
  
  def test_main(server):
      check_response(HOST, PORT)
      check_response_time(HOST, PORT)
      check_response(HOST, PORT)
      check_response_time(HOST, PORT)
  

Execution Result

$ curl http://localhost:8765
curl: (7) Failed to connect to localhost port 8765: Connection refused
$ pytest test_main.py
=============== test session starts =============== platform win32 -- Python 3.7.5, pytest-6.2.4, py-1.10.0, pluggy-0.13.1
rootdir: .
collected 1 item

test_main.py .                                                                                                                                                                                                                         [100%]

=============== 1 passed in 20.23s ===============
$ curl http://localhost:8765
curl: (7) Failed to connect to localhost port 8765: Connection refused
$ sed -i -e "s/WORKERS = 1/WORKERS = 3/g" test_main.py
$ curl http://localhost:8765
curl: (7) Failed to connect to localhost port 8765: Connection refused
$ pytest test_main.py
=============== test session starts =============== platform win32 -- Python 3.7.5, pytest-6.2.4, py-1.10.0, pluggy-0.13.1
rootdir: .
collected 1 item

test_main.py .                                                                                                                                                                                                                         [100%]

=============== 1 passed in 20.21s ===============
$ curl http://localhost:8765
"hello world"

$ # Why is localhost:8765 still alive?