This makes code written in Pandas 2.2.1 incompatible with 2.2.0. Was this done by mistake or on purpose? As far as I know, Pandas follows the semver methodology and this change is unexpected for me.