Shell executor not receiving SIGTERM from GitLab in Docker container

Summary

We have GitLab running inside a docker container on a host. On that same host, we've installed gitlab-runner and registered a shell executor runner with GitLab. All that works fine. However, when cancelling a job from the GitLab UI, no SIGTERM or SIGKILL is getting to the process, so job processes are being left behind, preventing subsequent jobs from kicking off.

Actual behavior

Job process is left behind

Expected behavior

Job process should be cancelled properly by SIGTERM.

Environment description

Version 13.10 gitlab/gitlab-ce docker container and version 13.10 gitlab-runner. Both are running on the same host, which is CentOS7.9. Docker version 1.13.1. The docker container has PORTS 0.0.0.0:80->80/tcp, 22/tcp, 0.0.0.0:443->443/tcp.

[[runners]]
name = "89"
url = "http://oururl"
token = "ourtoken"
executor = "shell"
output_limit = 8192