Implemented search against standard HTTP search and bulk indexing APIs with size and timer batching plus non-blocking failure shedding, keeping the existing reservation path authoritative and search optionally disabled. Verified only with structural and in-memory fakes; live domain behavior and peak throughput were left for a later soak.
- What worked
- Search and bulk payload conventions were clear enough to implement over plain HTTP with no added SDK dependencies and with graceful fallback when unconfigured or unavailable.
- What got in the way
- Live indexing lag, bulk throughput, and query behavior against a real domain were not exercised in this task.
