fix(tracking): restore trace persistence after merge

[RKS (rksharma-owg)]

Problem

main currently raises UnboundLocalError for every tracking trace because _add_trace() references the exception variable e on the successful path.

This breaks normal trace recording before the backend-enabled check and causes five existing unit tests to fail.

Root cause

PR #518 correctly added _record_failure(...) to the except block around backend trace persistence. When #518 was merged alongside #519's trace-lock changes, the call was misplaced immediately after _emit(...), outside the exception handler.

Implementation

Move _record_failure(...) back into the backend persistence except block. Successful local and persisted traces no longer reference an undefined exception, while real persistence failures continue to be recorded on the parent run.

Testing

• uv run ruff check .
• uv run ruff format --check hackagent/router/tracking/tracker.py
• uv run pytest -q tests/unit/attacks/test_error_propagation.py tests/unit/attacks/test_evaluation_updates.py tests/unit/router/tracking/test_goal_tracker_concurrency.py — 28 passed
• uv run pytest tests/ -q — 2,733 passed, 135 skipped, 177 subtests passed

Risk

Low. This is a one-line relocation that restores the behavior already reviewed and tested in #518. It does not change public APIs, schemas, or successful trace ordering.

Backward compatibility

Fully backward compatible.

Future improvements

Because the regression was introduced while merging two independently green PRs, running the unit-test gate against the final merge result would catch this class of conflict-resolution error before main advances.

[codecov]

Codecov Report

❌ Patch coverage is 0% with 1 line in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
hackagent/router/tracking/tracker.py	0.00%	1 Missing ⚠️

📢 Thoughts on this report? Let us know!

[Nicola Franco (franconicola)]

LGTM

[Copilot]

Pull request overview

Restores correct tracking trace persistence behavior by moving _record_failure(...) back into the backend persistence except block in the tracker, preventing UnboundLocalError on successful trace paths and preserving failure auditing when persistence actually fails.

Changes:

• Remove erroneous _record_failure(..., e) call from the non-exception (success) path in _add_trace().
• Record persistence failures only within the except Exception as e: block when backend.create_trace(...) raises.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.