[BUG]: ddtrace crashes containers without logs or stack trace (langgraph==1.0.4 + langchain==1.1.0 )

[fvalencia]

Tracer Version(s)

4.0.0, 3.17.1, 3.19.0

Python Version(s)

3.13.9

Pip Version(s)

N/A

Bug Report

We have identified an issue when using ddtrace that caused our containers to instantly die without logs or stack trace because of an OOM, the issue is not so easy to reproduce but every time it happens is during a langgraph graph execution, we have tried with different providers (OpenAI, Anthropic, GoogleAI) and the error happens with all of them.

Our application uses Fast API, and we recently upgraded to:

langchain==1.1.0
langgraph==1.0.4

Disabling ddtrace fixes the issue on our system.

Reference bugs that share a similar cause:

• ddtrace with FastAPI + OpenAI causes FastAPI container to instantly die with no logs or stack trace #6701
• Silent ddtrace crashes on 2.5 with FastAPI #8228

Reproduction Code

N/A

Error Logs

No error logs the containers die immediately.

Libraries in Use

datadog==0.51.0
ddtrace==4.0.0 # 3.17.1 or 3.19.0
langchain==1.1.0
langgraph==1.0.4

Operating System

Linux X86_64

[sabrenner]

Hi @fvalencia, thanks for opening this issue! I wonder if this is related to #15494. Maybe just to confirm/rule it out, the above issue is related to the wrapt dependency version, would you mind sharing what your wrapt version is?

It's possible that the above issue is not related, but I just want to see if we can rule it out. I'm not 100% sure because I feel like if it was related there would be some logs, which you aren't getting

[fvalencia]

Hi @sabrenner thanks for replying.

I upgraded to the latest version of ddtrace==4.1.2 and can confirm we have wrapt==2.0.1 and the server continues crashing when DD_PROFILING_ENABLED:true