Session Handoff [auto-2394]: Read the Closed Handoff List Once for a Closed Head per #2394 #2396

Next steps, in priority order

1. Fix handoff.py Reads the Closed Handoff List Twice for a Closed Head #2394 in scripts/handoff.py: where a track's head is closed, cmd_chain and cmd_new call newest_closed and then track_links, and each reads up to CLOSED_WINDOW closed handoff issues with their bodies. Derive the closed head from the Lane that track_links returns rather than from a second read, so the head and its head_doubts come from one snapshot and one request. Keep newest_closed's refusals exactly as they are: an unreadable row, and a full window refused only where no link on the track was found. Done looks like: one closed-list read per chain or new on a closed head (assert it with the test suite's FakeGh call record), the existing cases on those refusal messages in tests/test_handoff.py still passing unchanged, the new test failing with the fix reverted, the full test suite, the repo gate, the prose gate, and spec/validate.py passing, and a local-strict-review pass recorded at the pushed head.
2. Drive the pull request into develop with Closes on promotion: #2394 in its body, scope develop.

External blockers

None known.

Internal dependencies

Step 2 follows step 1. #2393 (on develop, not yet promoted) added the head_doubts round checks this read feeds, so base the work on current develop rather than main. #2395 (link skipping the round checks) touches the same file but leaves a choice open, so it is not part of this lane, and this change must not alter what link does.

State

No branch or worktree yet. Base the work on develop, per repo-worktree. Re-derive open pull requests touching scripts/handoff.py or tests/test_handoff.py before starting.

The parked decision queue

None for this lane.

What the last round did

Nothing yet. The unattended picker created this lane from the open backlog.

What not to repeat

Nothing recorded yet.

New learnings

None yet.