fix(event_bus): prevent GC of fire-and-forget tasks in fire_event()

[isharak7m]

Summary

fire_event() in src/event_bus.py creates asyncio tasks via loop.create_task() without storing the returned Task object. asyncio only keeps weak references to such tasks, so Python's GC can collect them before _handle_event completes — silently dropping the event handler with no error or log. This PR adds a _BG_TASKS set (matching the existing pattern in routes/chat_helpers.py:56-69 and src/builtin_mcp.py:93-105) that holds strong references until tasks finish.

Linked Issue

Fixes #6288

Type of Change

• Bug fix (non-breaking change that fixes an issue)

What Changed

• Added _BG_TASKS: set[asyncio.Task] = set() to src/event_bus.py
• Updated fire_event() to store the task in _BG_TASKS and register a done callback to discard it after completion

Why

fire_event() runs from 46 call sites across 16 files, triggering message_sent, session_created, document_created, email_received, memory_added, skill_added, and research_completed events. The returned Task was never stored — asyncio only keeps weak references to it, so the GC can collect and silently discard it before _handle_event completes. The codebase itself documents this hazard in app.py:1054-1057 and already uses the _BG_TASKS pattern in two other locations.

Diff

@@ -16,6 +16,12 @@ from src.constants import AUTH_FILE
 
 logger = logging.getLogger(__name__)
 
+# Strong references to fire-and-forget tasks. asyncio only keeps weak
+# references to tasks created via create_task(), so without this the GC can
+# collect a task mid-execution and the event handler silently never runs.
+# Mirrors _BG_TASKS in routes/chat_helpers.py and src/builtin_mcp.py.
+_BG_TASKS: set[asyncio.Task] = set()
+
 _task_scheduler = None
 
 
@@ -37,7 +43,9 @@ def fire_event(event_name: str, owner: Optional[str] = None):
     """
     try:
         loop = asyncio.get_running_loop()
-        loop.create_task(_handle_event(event_name, owner))
+        task = loop.create_task(_handle_event(event_name, owner))
+        _BG_TASKS.add(task)
+        task.add_done_callback(_BG_TASKS.discard)
     except RuntimeError:
         # No running loop — run in a new one (shouldn't happen in FastAPI)
         asyncio.run(_handle_event(event_name, owner))

How to Test

1. Run python -m pytest tests/test_event_bus_task_gc.py -v --noconftest — 4 tests should pass
2. Verify fire_event() now stores the task in _BG_TASKS and removes it after completion
3. Verify forced gc.collect() during handler execution does not lose the task

tests/test_event_bus_task_gc.py::TestEventBusTaskGC::test_task_survives_gc           PASSED
tests/test_event_bus_task_gc.py::TestEventBusTaskGC::test_task_removed_after_done     PASSED
tests/test_event_bus_task_gc.py::TestEventBusTaskGC::test_concurrent_events_tracked   PASSED
tests/test_event_bus_task_gc.py::TestEventBusTaskGC::test_no_loop_fallback            PASSED
======================== 4 passed in 0.60s ========================

Checklist

• I searched open issues and open PRs — this is not a duplicate.
• This PR targets dev
• My changes are limited to the scope described above — no unrelated refactors or whitespace changes mixed in.
• I actually ran the app and verified the change works end-to-end. Type-checks and unit tests are not enough.