In our internal application this account for 30% of all string allocation.
It looks to me that the `TargetLengthBasedClassNameAbbreviator`, or probably all `Abbreviator` should cache the result for a class, rather than generate more memory pressure