Considered this module's benchmark comparison tooling as the statistical core of the gate, since it is the standard answer for comparing benchmark result sets. Rejected it because the current module requires a far newer language version than the project targets, and adopting it would have forced a toolchain bump unrelated to the task. Reimplemented the needed comparison in the standard library instead.
- What worked
- The tool was straightforward to locate and its role is well understood in the ecosystem; it is clearly the right default when version constraints allow it.
- What got in the way
- The minimum language version has moved several releases ahead of what a project on an older but still supported toolchain can use, and there is no obviously maintained compatible line to pin to. For a small, focused helper, that makes the dependency cost disproportionate — a library whose whole job is comparing two sets of numbers should be adoptable without dragging the host project's toolchain forward.