[stefan-jansen]

After an order-state fault clears Alpaca connection health, public disconnect() returns without stopping the retained trading stream. Clean up retained stream/client resources regardless of the health flag, while preserving idempotent disconnect.

Closes #128.

Regression: uv run pytest -q tests/unit/test_alpaca_broker.py::test_disconnect_stops_the_stream_after_a_provider_order_fault.

• Producing base 3ddba2d: public connect and submit start a real adapter stream task. A validated SDK TradeUpdate reaches the registered handler and poisons order state. Public disconnect then fails the regression because the transport's stop method was called zero times.
• Fix 078b776: the unchanged workflow passes. Disconnect stops the transport, waits until its task is done, and a second disconnect does not stop it twice.
• Only external transports are replaced. The SDK model accepts the provider payload; the adapter's own cumulative-fill validation produces the fault. Test cleanup always releases its own blocking transport, including on the failing base.
• The final test was rerun against complete original producing source loaded from git archive. Three focused disconnect tests, Ruff check/format, ty and commit hooks passed. Full hosted qualification must pass before merge.

No release, deployment, live-money order or credentialed provider claim is included. A release requires matching Alpaca provider evidence for the changed adapter.

AI assistance: Codex reproduced the shutdown failure, implemented the cleanup correction, reviewed the diff and ran fail/pass verification.

[Copilot]

Copilot was unable to review this pull request because the user who requested the review has reached their quota limit.

[stefan-jansen]

Superseded by the combined maintainer-directed maintenance PR #121 to avoid qualifying and merging this queue separately. This PR head and its regression/documentation evidence are preserved in #121 and in a verified recovery archive. The accepted issue remains open until the combined fix lands; this closure does not claim the fix is merged. Follow-up work continues in #121.