Over the weekend I hacked up an example of what a fix could look like for this: https://github.com/chrisdennis/hibernate-orm/commits/HHH-10707

I wouldn't go so far as to call it a "proposal" but it at least proves it's fixable with two caveats - (that I'm aware of):

1. It appears the Infinispan tests do not cope well with the fixes in behavior - although the way they fail implies the tests are not providing coverage at the moment.
2. This will change the types inside the map returned from SecondLevelCacheStatistics.getEntries() (exposing any internal types used inside the strategy).