Weird behavior with follow=True blocking execution when running container exits

Environment:

OS: Ubuntu 22.04.1 LTS
Docker Engine: 20.10.21
Python version: 3.10.6
docker-py: 6.0.1

Context

The Python script below runs wrapped by a GitLab runner process that is a CLI application written with Typer.
The logging driver of Docker is journald

GitLab Runner: 15.6.1
Typer: 0.7.0

Question

I am trying to print logs from a running container. (it's execution is very fast)

import docker

cli = docker.from_env()
container = cli.containers.run(image=image, detach=True, remove=False)

for cl in container.logs(stream=True, follow=True):
    typer.secho(cl.strip(), fg=typer.colors.BLUE)

container.remove()

For a reason, container.logs(stream=True, follow=True) is acting weirdly whenever the detached container exits. Sometimes the execution of my program resumes and finishes properly and very often, the underlying container exits and my program get stuck in the for loop, doing nothing.

This has for consequence that jobs running in GitLab are timeouting due to no exit code of my cli.

It seems most likely system related since I can run the cli from my laptop without trouble. Any ideas from where the problem could come?

Start with the container.logs(stream=True, follow=True) call in the provided script and reproduce the behavior with Docker Engine 20.10.21, journald logging, and a short-lived detached container. Compare the stream behavior after the container exits and confirm that the CLI loop terminates and container.remove() is reached.