Factor the benchmark scenarios into standalone example binaries that can be traced directly without cargo run, and simplify the hot call and hook lookup paths using the benchmark output as guidance.
Measure representative protocol paths in nanoseconds over hundreds of samples and report standard deviation so framing, routing, and local-delivery changes can be compared with less jitter.
Implement an aligned two-section frame format, a compiled prefix router, a minimal pending and active hook engine, and a header-first receive path that only decodes payloads on local delivery. Recreate the protocol-focused test suite and document the explicit framing deviation in src/protocol/PROTOCOL_CHANGES.md.
Compile routing prefixes once per endpoint, restore minimal pending-to-active hook transitions, and route call/data/fault packets from the header before decoding payloads for local delivery only. Document the remaining protocol-level pressure points in src/protocol/PROTOCOL_CHANGES.md.
Remove the unused pending-hook layer and dead protocol trait wrappers, and resolve active hooks through a direct active-only flow with peer-side indexing instead of scan-based recovery.
Replace vector-backed endpoint outcomes with single-action results, skip payload deserialization on forwarded data and faults, and route local call and data emissions without encode/decode roundtrips.
Move demo leaf echo behavior out of the core protocol runtime, treat procedure IDs as opaque protocol fields, and return direct registered children in endpoint introspection to match the spec.
Document public APIs and non-obvious control flow so the protocol, simulator, and macro crates are easier to follow. Tighten a few helper paths and feature gates while preserving behavior and keeping the workspace warning-free.
Michael Mikovsky