Warm up after half a tick and double the worker target while probes pay off, dropping to AIMD +2 on first plateau/error burst. 33% faster in shaped benchmarks; backoff semantics unchanged.
Opt-in --adaptive-workers (AIMD scaling + tail range-split), request/rnet session parity, and fixes for resume poisoning, truncated-segment acceptance, 416 handling, and byte-range-slice corruption. Tests + docs.