After an error, it is possible for a storage server process to recruit a second storage instance

We've observed a situation where a storage server encountered an `unknown_error` that terminated the role. About a minute later, a new storage server was recruited on this process without the process having ever died.

As a result the storage server process had two sets of storage files and was consuming a lot more disk space. The process did not do anything to deal with the old storage files, though it likely would have deleted them upon restart assuming the data in them had been re-replicated.