- Batch query speed up to 3× by reading each point once; Gridstore migration improves storage performance and consistency.
- Added configurable 60 s internal timeout for telemetry endpoints and user‑configurable inference request timeout; validated shard‑transfer restarts and deferred Gridstore flushing.
- Fixed multiple crashes: startup panic on old clusters, Raft snapshot crash loop, WAL corruption, data race in shard transfers, and various logging/segment issues.