v5.2.1

Performance

• Read-ahead, async prefetch, and range cache for tunneled fio reads (#24) —
  the server-side fio layer now batches reads, prefetches the next block while
  the current one is consumed, and caches recently-read ranges, cutting
  request/response round trips for readers like FFmpeg's AVIO layer that pull
  ~32 KiB at a time. Reduces over-IP overhead on sequential workloads from
  roughly 8–14× slower than local to under 2×. Tunable via the
  FFOIP_READAHEAD_BYTES and FFOIP_RANGE_CACHE_BYTES environment variables
  (set FFOIP_READAHEAD_BYTES=0 to disable). No protocol or configuration
  changes are required.

Full Changelog: v5.2.0...v5.2.1