Resolves #11285

Authored by Claude Opus 4.7 (Claude Code 1M context).

Evidence: L1 (10 unit tests covering all dual-search paths) → L1 required (no runtime ACs; file-system fixture coverage). No residuals.

What shipped

B2 of Epic #11187 7-bucket fan-out — LocalFileService read-path dual-search during active-to-archive migration transition.

getIssueById changes:

• Tries new canonical archiveRoot first
• Falls back to legacy archiveDir if not found in archiveRoot
• Once Epic #11187 B1 corpus migration completes, legacy fallback becomes a cheap miss

getDiscussionById changes:

• Tries new flat path (discussionsDir/discussion-N.md) first — post-B1 canonical shape
• Falls back to recursive XXxx subdir search in discussionsDir — covers pre-B1 active shape (currently 23 legacy XXxx subdirs + 23 new discussion-XXxx subdirs per OQ1 duplication state)
• Then searches archiveRoot recursively — no legacy discussion-archive substrate ever existed per Epic #11187 body

Bonus substrate-correctness fix: before this change, getDiscussionById could not find ANY active discussion under the current legacy XXxx subdir shape (it only tried flat root). The dual-search now correctly resolves the existing discussion corpus.

AC Coverage (#11285)

• AC1: LocalFileService#getIssueById checks both old (archiveDir) and new (archiveRoot) paths. ✓

Out-of-scope per ticket OoS:

• Write-path / syncer logic unchanged
• getPullById not added — does not currently exist in LocalFileService; ticket says "(and PR equivalent)" but precedent doesn't establish the method shape. Defer to follow-up if PR read-path needed.

Test Evidence

• npm run test-unit -- test/playwright/unit/ai/services/github-workflow/LocalFileService.spec.mjs → 10/10 pass (524ms)
• npm run test-unit -- test/playwright/unit/ai/services/github-workflow/ → 80/80 pass (2.6s) — no regression in sibling test suites
• git diff --check origin/dev...HEAD clean

Test scenarios

1. getIssueById finds active issue via chunkPath shape
2. getIssueById falls back to new archiveRoot when not in active
3. getIssueById falls back to legacy archiveDir when not in archiveRoot
4. getIssueById returns NOT_FOUND when issue exists nowhere
5. getIssueById prefers archiveRoot over archiveDir when both have the same id (ordering)
6. getDiscussionById finds active flat discussion (post-B1 canonical shape)
7. getDiscussionById falls back to legacy XXxx subdir (pre-B1 active shape)
8. getDiscussionById falls back to archiveRoot when not in active
9. getDiscussionById returns NOT_FOUND when discussion exists nowhere
10. getIssueById accepts leading # in issue number

Test substrate discipline

describe.serial applied to prevent fullyParallel races on shared aiConfig mutation per feedback_symmetric_spec_cleanup memory anchor (Playwright interleaves tests in same worker; cross-singleton config mutations need serial ordering for fixture file lifecycle).

Epic #11187 context

• B0b (MetadataManager pruning fix) ✓ MERGED via PR #11282
• B1 (#11284 — Gemini): active discussions collapse + issue-archive migration
• B2 (this PR, #11285): LocalFileService dual-search ← lands BEFORE B1 (no-op for current substrate; ready for post-B1 reads)
• B3 (#11286 — me): data-sync-pipeline.yml + buildScripts archive substrate
• B4 (#11287): 11-file consumer cleanup (TicketSource, SEO, analyzers)
• B5 (#11288): tests + AC16/17 + anomaly hook
• B0a + AC8: NOT YET FILED (surfaced in MESSAGE:ed9e146f-... — awaits Gemini/GPT filing)

Cross-Family Review Routing

Primary-reviewer: @neo-gemini-pro per cross-family rotation (B1 substrate-author on adjacent corpus migration). A2A handoff with commentId follows.

Operator merge gate

@tobiu — per AGENTS.md §0 Invariant 1, merge reserved exclusively for you. Eligible for human merge once CI green + cross-family approves.

🤖 Generated with Claude Code