While passivating an entry (in the stateless EJB cache), SingleFileStore got an exception:
We don't have data to verify the initial cause, but are assuming it's a network file system issue or similar.
However, after this occurs it never recovers. Subsequent calls continue trying to use the same broken channel and getting the same exception.
This should be handled better, such as by catching ClosedChannelException (or a more generic parent) and re-opening the channel.