max-request, reload-on-rss are applicable only for workers. Then how to avoid memory leaks in mules?

I’m observing continuous anonymous memory growth in a uWSGI mule process in a Python application.

According to the official uWSGI documentation, options such as max-requests and reload-on-rss are available to mitigate memory leaks in worker processes, but these settings do not apply to mule processes.

Given this limitation, what are the recommended approaches to handle or mitigate memory growth in uWSGI mule processes?
Are there any supported uWSGI configurations, design patterns, or best practices (such as recycling mules, isolating workloads, or externalizing tasks) to address this issue?

Any guidance or real-world solutions would be appreciated.