In a previous blog post, I described some benchmarks I wrote for a program written in Rust.While presenting the results, I mentioned a strange behavior: things that should have been very fast (a few nanoseconds) were reported as instantaneous.I wrote that it was probably fine given that the bigger benchmarks (above 10 nanoseconds) were seemingly working well, following the expected asymptotic behavior.However, this prompted me to dig deeper to understand what was going on under the hood.