fix(status): hide stale incomplete-turn recovery after workspace cleanup

[moncher-dev]

Problem

gh-symphony repo status can keep reporting a "Recoverable incomplete turn" after the associated issue workspace has already been removed.

Observed local state:

• status.json still contains recovery.kind = "incomplete-turn-dirty-workspace".
• The recovery workspacePath points to a deleted repository directory.
• The run record is still preserved with status = "suppressed" and recovery != null.
• The issue workspace record has status = "removed" and its repositoryPath no longer exists.
• gh-symphony repo recover --dry-run --json returns [], so the status output advertises a recovery that the recovery command cannot act on.

This makes repo status look actionable even though the recovery context is stale.

Root cause

This appears to be an Observability Layer bug crossing the Execution Layer workspace lifecycle boundary:

• cleanupTerminalIssueWorkspace() removes the issue workspace and marks the workspace record as removed, while intentionally preserving run records.
• buildProjectSnapshot() derives snapshot.recovery from allRuns only.
• findLatestRecovery() suppresses stale recovery after a later recovery run completes, but it does not know whether the workspace backing the recovery has since been removed.
• repo status renders snapshot.recovery without validating that the workspace still exists or is still active.

Proposed fix

Prefer keeping run records immutable/preserved, and make the status snapshot smarter:

1. Pass issue workspace records into buildProjectSnapshot() or otherwise provide a resolver for recovery workspace state.
2. Treat an incomplete-turn-dirty-workspace recovery as resolved/stale when its issue workspace record is removed.
3. Optionally also suppress the recovery when the recorded recovery workspacePath no longer exists.
4. Keep the run record intact for audit/history.

This keeps the cleanup behavior intact while making the Observability Layer report only currently actionable recovery.

Expected behavior

After a terminal cleanup removes an issue workspace:

• gh-symphony repo status should not show "Recoverable incomplete turn" for that removed workspace.
• status.json.recovery should be null.
• Historical run records may still preserve their recovery payload.
• repo explain should not warn that dispatch will start a recovery turn for a removed workspace.

Test cases

• Unit: buildProjectSnapshot() does not surface recovery when the matching workspace record has status: "removed".
• Unit: buildProjectSnapshot() still surfaces recovery when the workspace record is active and dirty recovery is unresolved.
• Orchestrator/service: create a suppressed incomplete-turn dirty-workspace run, then run terminal workspace cleanup; the saved project status should have recovery: null.
• CLI: repo status does not render the "Recoverable incomplete turn" block when snapshot recovery is null.

Notes

This should not modify docs/symphony-spec.md. The change aligns status reporting with the existing workspace lifecycle instead of changing the upstream spec.

[moncher-dev]

Workpad — #421 — Cycle 1

Type: fresh pickup
Branch: feat/421-hide-stale-recovery
Draft PR: not yet created
Cycle opened: 2026-07-02T16:07:31Z · Trigger: Ready fresh pickup for stale incomplete-turn recovery status fix

Status Transitions

Plan

• 1. Implement recovery suppression for removed/missing workspaces with unit/service/CLI coverage
• 2. 마무리: 원본 이슈 재검증 · Completion Bar 통과 · changeset (필요 시) · PR ready · In review 전이

Rework / PR Feedback

• Major merge blockers from review:
  • None

Completion Bar

• In-scope requirements implemented
• pnpm lint
• pnpm test
• pnpm typecheck
• pnpm build
• Docker E2E (if integration behavior changed)
• Tests for new functionality (or justified N/A)
• Code conventions (CLAUDE.md)
• Inline review comments answered (rework only)

Validation

• Pending
• Changeset: pending
• Docker E2E evidence: pending
• Merge commit (Land cycle only): N/A

위임 (out-of-scope / human / post-merge)

• None

Progress Log

• 2026-07-02T16:07:31Z: Opened Cycle 1 for Ready fresh pickup; classified scope as Observability Layer crossing Execution Layer workspace lifecycle.

Blockers

None

[moncher-dev]

Workpad — #421 — Cycle 1

Type: fresh pickup
Branch: feat/421-hide-stale-recovery
Draft PR: #426
Cycle opened: 2026-07-02T16:07:11Z · Trigger: Fresh Ready pickup; no linked/open PR or actionable review feedback found.

Status Transitions

• 2026-07-02T16:08:13Z · Ready → In progress · Fresh pickup triaged as actionable; draft PR fix(status): hide stale cleanup recovery #426 opened for implementation. (cycle 1 open)
• 2026-07-02T16:18:37Z · In progress → In review · Completion Bar passed; PR fix(status): hide stale cleanup recovery #426 ready for human review. (cycle 1 close)

Plan

• 1. Implement stale recovery suppression in the project status snapshot using issue workspace lifecycle state, with targeted unit/service/CLI coverage.
• 2. Finish: original issue recheck · Completion Bar pass · changeset if required · PR ready · In review transition

Rework / PR Feedback

N/A — fresh pickup.

Completion Bar

• In-scope requirements implemented
• pnpm lint
• pnpm test
• pnpm typecheck
• pnpm build
• Docker E2E (if integration behavior changed)
• Tests for new functionality (or justified N/A)
• Code conventions (CLAUDE.md)
• Inline review comments answered (rework only; N/A)

Validation

• Layer classification: Observability Layer fix crossing Execution Layer workspace lifecycle boundary; docs/symphony-spec.md unchanged.
• Spec conformance: aligns status reporting with existing workspace lifecycle; no intentional upstream spec divergence.
• pnpm lint — pass
• pnpm test — pass
• pnpm typecheck — pass
• pnpm build — pass
• pnpm exec prettier --check .changeset/hide-stale-cleanup-recovery.md packages/cli/src/commands/repo-explain.ts packages/cli/src/commands/status.test.ts packages/core/src/observability/snapshot-builder.test.ts packages/core/src/observability/snapshot-builder.ts packages/orchestrator/src/dispatch.test.ts packages/orchestrator/src/explain.ts packages/orchestrator/src/service.test.ts packages/orchestrator/src/service.ts — pass
• pnpm format — fail on existing unrelated repository formatting drift, including docs/symphony-spec.md; changed-file Prettier check passed.
• Targeted TC: buildProjectSnapshot() suppresses removed-workspace recovery and preserves active-workspace recovery; service cleanup preserves run recovery but saves status recovery null; CLI status hides block when snapshot recovery is null; repo explain no longer warns for removed workspace.
• Docker E2E: docker compose -f docker-compose.e2e.yml -f docker-compose.e2e.events.yml up -d --build; health passed; happy-path fixture status API showed health: running, activeRuns: 1, dispatched: 1, recovery: null.
• Changeset: .changeset/hide-stale-cleanup-recovery.md
• Docker E2E evidence: evidence/issue-421-docker-state.json
• Merge commit (Land cycle only): N/A

Delegation (out-of-scope / human / post-merge)

• None

Progress Log

• 2026-07-02T16:07:11Z: Opened cycle 1 after Ready triage; no existing linked/open PR found.
• 2026-07-02T16:08:13Z: Created draft PR fix(status): hide stale cleanup recovery #426 and transitioned the issue to In progress.
• 2026-07-02T16:16:39Z: Implemented workspace-lifecycle-aware recovery filtering, repo explain suppression, regression tests, and patch changeset.
• 2026-07-02T16:18:37Z: Passed validation, marked PR fix(status): hide stale cleanup recovery #426 ready, and transitioned issue to In review.

Blockers

None

[moncher-dev]

🔁 Status: Ready → In progress

Reason: Fresh pickup triaged as actionable; draft PR #426 opened for implementation.
Cycle: 1 open

[moncher-dev]

🔁 Status: In progress → In review

Reason: Completion Bar passed; PR #426 ready for human review.
Cycle: 1 close