docs(09-01): complete retention cleanup plan

- Create 09-01-SUMMARY.md with execution results
- Update STATE.md with phase 9 position and decisions
- Update ROADMAP.md with phase 9 progress
- Mark STOR-03 and STOR-04 requirements complete

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

.planning/phases/09-retention-cleanup/09-01-SUMMARY.md

@@ -0,0 +1,98 @@
+---
+phase: 09-retention-cleanup
+plan: 01
+subsystem: database
+tags: [apscheduler, retention, postgresql, prometheus, cascade-delete]
+
+# Dependency graph
+requires:
+  - phase: 01-database-schema
+    provides: router_config_snapshots table with CASCADE FK constraints
+provides:
+  - Automatic retention cleanup of expired config snapshots
+  - CONFIG_RETENTION_DAYS env var for configurable retention period
+  - Prometheus metrics for cleanup observability
+affects: []
+
+# Tech tracking
+tech-stack:
+  added: []
+  patterns: [APScheduler IntervalTrigger for periodic maintenance jobs]
+
+key-files:
+  created:
+    - backend/app/services/retention_service.py
+    - backend/tests/test_retention_service.py
+  modified:
+    - backend/app/config.py
+    - backend/app/main.py
+
+key-decisions:
+  - "make_interval(days => :days) for parameterized PostgreSQL interval (no string concatenation)"
+  - "24h IntervalTrigger with 1h jitter to stagger cleanup across instances"
+  - "AdminAsyncSessionLocal (bypasses RLS) since retention is cross-tenant system operation"
+
+patterns-established:
+  - "IntervalTrigger pattern for periodic maintenance jobs (vs CronTrigger for scheduled backups)"
+
+requirements-completed: [STOR-03, STOR-04]
+
+# Metrics
+duration: 2min
+completed: 2026-03-13
+---
+
+# Phase 9 Plan 1: Retention Cleanup Summary
+
+**Daily APScheduler job deletes config snapshots older than CONFIG_RETENTION_DAYS (default 90) with CASCADE FK cleanup of diffs and changes**
+
+## Performance
+
+- **Duration:** 2 min
+- **Started:** 2026-03-13T04:31:48Z
+- **Completed:** 2026-03-13T04:34:12Z
+- **Tasks:** 2
+- **Files modified:** 4
+
+## Accomplishments
+- Retention service with parameterized SQL DELETE using make_interval for safe interval binding
+- APScheduler IntervalTrigger running every 24h with 1h jitter for stagger
+- Prometheus counter and histogram for cleanup observability
+- Wired into main.py lifespan with non-fatal startup pattern
+
+## Task Commits
+
+Each task was committed atomically:
+
+1. **Task 1 (RED): Add failing tests** - `00bdde9` (test)
+2. **Task 1 (GREEN): Implement retention service + config setting** - `a9f7a45` (feat)
+3. **Task 2: Wire retention scheduler into lifespan** - `4d62bc9` (feat)
+
+## Files Created/Modified
+- `backend/app/services/retention_service.py` - Retention cleanup logic, scheduler, Prometheus metrics
+- `backend/tests/test_retention_service.py` - 4 unit tests for cleanup function
+- `backend/app/config.py` - Added CONFIG_RETENTION_DAYS setting (default 90)
+- `backend/app/main.py` - Wired start/stop retention scheduler into lifespan
+
+## Decisions Made
+- Used make_interval(days => :days) for parameterized PostgreSQL interval (avoids string concatenation SQL injection risk)
+- 24h IntervalTrigger with 1h jitter to stagger cleanup across instances
+- AdminAsyncSessionLocal bypasses RLS since retention is a cross-tenant system operation
+
+## Deviations from Plan
+
+None - plan executed exactly as written.
+
+## Issues Encountered
+None
+
+## User Setup Required
+None - no external service configuration required. CONFIG_RETENTION_DAYS defaults to 90 if not set.
+
+## Next Phase Readiness
+- Retention cleanup is fully operational, ready for phase 10
+- No blockers
+
+---
+*Phase: 09-retention-cleanup*
+*Completed: 2026-03-13*