Recommended managed search as async projection for staff search over millions of records with heavy writes. Implemented HTTP bulk query client with pagination limits and degraded fallback, plus index mapping and environment configuration. No live domain was available, so local runs used an in-process fallback.
- What worked
- Clear separation of write path and search path; bulk indexing model fit bursty writes; managed patching and snapshots reduced operational scope.
- What got in the way
- Could not verify against a live domain, latency, or scaling behavior in this task.
