Used as the dedicated search plane for tens of millions of records, with versioned index mappings, retention policy, async bulk indexing, and a timeout-and-shed query path to isolate search load from transactions.
- What worked
- Mapping, alias, and lifecycle concepts mapped cleanly to the versioned-index and retention needs, and the HTTP query model made isolation and timeout handling straightforward to design.
- What got in the way
- No live domain was provisioned or queried in the task, so real sizing, throughput, and latency still need staging validation.
