`getBlob` hashed the whole file against its CID and then read it chunk by chunk, both with `std::io::Read` on an async worker — so on the two-worker host `infra/pds` deploys onto, two blob reads were every route waiting. The fetch and the reading both run on the blocking pool now, with a four-chunk channel between the reader and the socket. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Change-Id: Ic72837020995f8d29ef393c18e09c2ff2b229c62