Hi Niklas, thanks for the response... we double checked our "socket footprint": it seems to be perfectly normal (a little bit more than the number of client threads, less than 32).
The second finding is that the repro tends to be more effective when you run the repro client -- which only issues LISTs -- against a directory preloaded with a several hundreds files. I'd say that this is not the same thing you experienced then... Any clues? Cheers, F. -- http://issues.apache.org/jira/browse/FTPSERVER-359