ManagedDMRContentTypeResource.content is used for generating an overall hash for the stored content, but it does not use consistent ordering (i.e. it needs to be a LinkedHashMap or perhaps a TreeMap.) The result is when another node in the domain receives an update it may calculate a different overall hash.

This is surfacing as failures in https://bugzilla.redhat.com/show_bug.cgi?id=1078062 when the master and slave are running on different JVM releases. Different VM releases often have different ordering behavior when iterating over the unordered collections.

I believe this is unlikely to result in real-world problems since any backup will persist its own version of the hash to its local copy of domain.xml, so there won't be any mismatch. The importance of the overall hash is that it allows the process to find the content in the repo when it is added.