fix(mcp): clear running loop state before creating background thread event loop

[strands-agent]

Description

Fixes #1512

This PR fixes the issue where MCPClient fails to start in ASGI environments (uvicorn/hypercorn) with the error:

RuntimeWarning: coroutine 'MCPClient._async_background_thread' was never awaited
MCPClientInitializationError: background thread did not start in 30 seconds

Root Cause

PR #1444 added contextvars.copy_context() to propagate context variables to the background thread (fixing #1440). However, in ASGI environments where the parent thread already has an active event loop, copying the context also copies the _running_loop contextvar marker.

When the background thread calls run_until_complete(), Python's asyncio detects this inherited running loop marker and raises:

Cannot run the event loop while another loop is running

Credit to @jpvelasco for identifying the actual root cause: the issue is the _running_loop contextvar (via asyncio.events._set_running_loop), not just the thread-local event loop.

Solution

Clear both the running loop marker and thread-local event loop before creating the new event loop:

def _background_task(self):
    # Clear the inherited "running loop" contextvar
    try:
        import asyncio.events as events
        events._set_running_loop(None)
    except (AttributeError, RuntimeError):
        pass
    
    # Also clear thread-local event loop  
    try:
        asyncio.set_event_loop(None)
    except RuntimeError:
        pass
    
    self._background_thread_event_loop = asyncio.new_event_loop()
    asyncio.set_event_loop(self._background_thread_event_loop)
    self._background_thread_event_loop.run_until_complete(self._async_background_thread())

Why This Works

1. ✅ Preserves context variable propagation from PR fix(mcp): propagate contextvars to background thread #1444 (auth and other context vars still work)
2. ✅ Clears inherited running loop marker - the actual root cause identified by @jpvelasco
3. ✅ Isolates event loops between parent and background threads
4. ✅ Defensive - handles partial initialization gracefully with try/except
5. ✅ No breaking changes - existing MCP functionality unchanged
6. ✅ Works with uvicorn/ASGI environments (verified by issue reporter)
7. ✅ Compatible with OpenTelemetry instrumentation

Related Issues

• Fixes [BUG] MCPClient background thread fails to start under uvicorn/ASGI - "coroutine was never awaited" #1512 (MCPClient fails in uvicorn/ASGI)
• Related to [BUG] MCP Client does not propagate contextvars #1440 (context propagation - preserved by this fix)
• Related to PR fix(mcp): propagate contextvars to background thread #1444 (added context copying that triggered this issue)

Documentation PR

No documentation changes required - this is an internal implementation fix.

Type of Change

Bug fix (non-breaking change which fixes an issue)

Testing

• Existing MCP tests pass (no regression)
• Fix addresses the exact reproduction case from issue [BUG] MCPClient background thread fails to start under uvicorn/ASGI - "coroutine was never awaited" #1512
• Solution verified by @jpvelasco (issue reporter)

Checklist

• I have read the CONTRIBUTING document
• I have added any necessary tests that prove my fix is effective or my feature works
• I have updated the documentation accordingly
• I have added an appropriate example to the documentation to outline the feature, or no new docs are needed
• My changes generate no new warnings
• Any dependent changes have been merged and published

---

By submitting this pull request, I confirm that you can use, modify, copy, and redistribute this contribution, under the terms of your choice.

---

PR created by strands-coder autonomous agent with fix verified by @jpvelasco 🧬