[Perf] - Cut memory leak in half

[AlexsanderHamir]

Title

[Perf] - Cut memory leak in half

Relevant issues

Addresses #12685

Pre-Submission checklist

Please complete all items before asking a LiteLLM maintainer to review your PR

• I have Added testing in the tests/litellm/ directory, Adding at least 1 test is a hard requirement - see details
• I have added a screenshot of my new test passing locally
• My PR passes all unit tests on make test-unit
• My PR's scope is as isolated as possible, it only solves 1 specific problem

Type

🧹 Refactoring

Changes

• Remove references from being passed to fire-and-forget tasks.

Results

• The memory leak is now reduced by more than half.
• There's still one small memory leak remaining

Before

After

How to look for memory leaks ?

1. Start the proxy server with a single worker.

2. In a separate terminal, run:

   memray attach --trace-python-allocators <ProxyServer's PID>

3. Run a load test against the proxy server.

4. Stop the load test.

5. Let the process sit for 5–10 minutes and observe what memory remains.

6. Sort by % Own

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Preview	Comments	Updated (UTC)
litellm	Ready	Preview	Comment	Dec 10, 2025 8:31pm