[AliAbouelazm]

Fixes #15413.

Use asgiref.local.Local when available so Django request context is isolated between async tasks and follows sync/async adapters. Keep the threading.local fallback without adding a dependency.

Add regressions for concurrent tasks, both adapter directions, ordinary thread isolation, and the optional-import fallback. Isolate existing request-data tests without clearing the storage object's internals.

Validation:

• 1,386 unit tests pass on Python 3.12 with both protobuf implementations.
• Ruff and Flake8 pass. Changed middleware and new tests have 100% line and branch coverage; aggregate coverage does not decrease from the baseline.
• Real ASGI/WSGI async-view probes pass on Django 5.2 and 6.1.
• No live Google Cloud resources were used.

AI-assisted implementation and validation.

[gemini-code-assist]

Code Review

This pull request updates the Django request middleware to use asgiref.local.Local when available, enabling support for both thread-local and async-local contexts, while falling back to threading.local if asgiref is not installed. It also adds comprehensive unit tests for async and thread-local behaviors. Feedback suggests refactoring a thread-local test to group assertions and middleware execution within a single worker function to prevent potential test flakiness from thread reuse assumptions in ThreadPoolExecutor.

[daniel-sanche]

Changes LGTM, but it looks like there are some failing tests to resolve

[daniel-sanche]

If it;s possible to use contextvars directly for this instead, that would be ideal. We try to avoid external dependencies where possible

But that said, this file is specifically for django middleware, and IIUC, django requires asgiref. And you handled the fall-back case. So I think this is fine as-is