[Task]: GH-C91 typed replan semantic-delta negative fixture

[liubf21]

Task ID or area

GH-C91 / control-plane replan semantics

Intent

I want to claim an existing task.

Summary

Add one semantics-derived negative regression for typed replan semantic-delta settlement, preserving the distinction between genuine successor progress and non-novel or ungrounded state changes.

Proposed scope

In scope:

• choose one contract-defined negative transition from the task board (equivalent coverage fingerprint, evidence-only churn, ungrounded successor id, repeated blocker, or stale legacy ACK);
• derive expected behavior from the typed contract rather than current implementation output;
• add focused pytest coverage and retain the existing public smoke seam.

Out of scope:

• changing default replan behavior, quota policy, scheduler cadence, or authority boundaries.

Relevant files or commands

• tests/control_plane/test_replan_semantic_action_behavior.py
• tests/control_plane/test_replan_novelty_policy.py
• examples/control_plane/quota-replan-decision-plane-smoke.py
• relevant status/evidence protocol docs

Validation plan

• focused replan pytest suite;
• python3 examples/control_plane/quota-replan-decision-plane-smoke.py;
• loopx check on touched status/evidence contracts and CONTRIBUTOR_TASKS.md.

Public/private boundary

• This issue contains no private benchmark traces, verifier output, credentials, internal document links, raw agent sessions, or local runtime state.
• The fixture will use only synthetic typed state and evidence.

[liubf21]

Baseline check found that the current GH-C91 row appears already satisfied on main; I am pausing rather than duplicating a semantic fixture.

Current coverage includes:

• evidence-only churn and repeated coverage scope: tests/control_plane/test_progress_observation.py (test_new_probe_family_is_a_semantic_delta_but_new_evidence_alone_is_not, test_terminal_coverage_requires_a_semantically_new_scope);
• repeated blocker: test_repeated_blocker_cannot_close_replan in the same file;
• stale rotated obligation ACK: tests/control_plane/test_replan_novelty_policy.py::test_semantic_ack_is_bound_to_the_exact_rotated_obligation and tests/control_plane/test_refresh_state_replan_gate.py::test_rotated_vision_obligation_rejects_first_maintenance_writeback;
• ungrounded/wrong successor binding: test_refresh_state_replan_gate.py::test_completed_or_wrong_obligation_successor_cannot_close_rotated_duty plus the end-to-end CLI rejection tests in test_replan_semantic_action_behavior.py.

Could a maintainer identify a remaining invariant that lacks coverage, or mark GH-C91 done/superseded on the contributor board? No implementation changes were made.

[huangruiteng]

mark GH-C91 done/superseded on the contributor board 吧

[liubf21]

Maintainer-directed board cleanup is in #3286.

[liubf21]

The maintainer-directed cleanup now continues in fork-based PR #3287.