fix: handle Python 3.12+ get_event_loop RuntimeError in VirtualKernelContext

[maartenbreddels]

Summary

Unit tests on master are failing on Python 3.12 across macOS and Ubuntu with:

RuntimeError: There is no current event loop in thread 'MainThread'.

Every test after tests/unit/starlette_lifespan_test.py (~30% of the suite) errors out at setup of the autouse kernel_context fixture.

Why

tests/unit/starlette_lifespan_test.py (added in 1a19f8a) calls asyncio.run(...). On Python 3.12+, asyncio.run() cleans up by calling set_event_loop(None), which puts the event-loop policy into a state where a subsequent asyncio.get_event_loop() call from the main thread raises RuntimeError instead of silently creating a new loop.

VirtualKernelContext.event_loop used asyncio.get_event_loop as its default_factory, so every test after the lifespan test failed when the kernel_context fixture constructed a new VirtualKernelContext.

Fix

Wrap the factory in a small helper that falls back to new_event_loop() + set_event_loop() when get_event_loop() raises. This matches the pre-3.12 auto-create behavior that the rest of the code was implicitly relying on.

Reproducer on 3.12:

import asyncio
async def dummy(): pass
asyncio.run(dummy())
asyncio.get_event_loop()  # RuntimeError: There is no current event loop in thread 'MainThread'

Test plan

• CI unit tests pass on Python 3.12 (previously failed)
• CI unit tests continue to pass on Python 3.8

🤖 Generated with Claude Code