HaloLog and the Problem With a 23.9 ns Benchmark
What happened
How HaloLog reached 23.9 ns/op, why a Zap maintainer’s review changed the harness, and how zero-allocation claims became executable tests.
Summary assembled by rule from the sources below