Re-fetch session snapshot before post-run event assertions

[davidahmann]

Problem

Post-run_async assertions that inspect a previously held in-memory session object can read stale event state, leading to false negatives and flaky test/operator checks.

Why now

This stale-snapshot pattern is recurring friction in runner/session assertions and can misdiagnose persistence behavior.

Evidence packet

• Commit under test: f27a9cfb87caecb8d52967c50637ed5ad541cd07 (origin/main, 2026-02-19)
• Runtime environment: Python 3.14.0, macOS 26.3
• Minimal repro:
  1. Create a session and retain the returned object.
  2. Execute runner.run_async(...) and append events.
  3. Assert counts from the original retained session object.
• Expected: assertions reflect persisted post-run event state.
• Actual: retained session objects may be stale; fresh get_session(...) is required to observe authoritative state.

Acceptance

• Update runner/session tests to assert event counts using freshly fetched session snapshots after run completion.
• Add/adjust regression coverage documenting stale-object vs re-fetch behavior.

[ryanaiagent]

Hi @davidahmann , Thank you so much for catching this and providing the test case. Your suggested PR is currently under review.