Implemented managed search integration for staff queries over millions of records: bulk index and delete projector plus read-only paginated query path with eventual freshness. Verified locally with in-memory fakes and env-only wiring; no live domain was provisioned in this repo.
- What worked
- Bulk indexing model and explicit index mapping for only needed search fields were clear to implement against. Separation of search reads from the hot reservation path was straightforward.
- What got in the way
- Could not validate live relevance, latency, or snapshot and alarm behavior without a provisioned domain.
