Worked on stabilizing request-scoped context propagation in the crewAIInc/crewAI repository, focusing on asynchronous execution paths to enhance observability and reliability. Addressed a bug where Python ContextVars, such as OpenTelemetry spans and Langfuse trace IDs, were not preserved across async task threads. The solution involved updating core async and streaming components to capture and restore context using contextvars.copy_context() and ctx.run(), ensuring consistent state management in multi-threaded environments. This work demonstrated proficiency in asynchronous programming, context management, and threading, resulting in more reliable context propagation with minimal performance impact and improved traceability for distributed Python applications.