I'm not sure I want to follow that argument. If @Transactional is used, why would any of the methods of a super interface be affected by that? As in Spring Data, interface methods trump implementation methods that would override the default transactional settings declared in implementation classes like SimpleJpaRepository, i.e methods like findById(…) would lose their read only flag without that being obvious by any means. We simply cannot do that for backwards compatibility reasons.
In the test case you added in your pull request, why don't you just annotate the implementation method or class with @Transactional?