This can occur if there is a random number collision, which is actually not as unlikely as it sounds. Because Random() uses a volatile and not an Atomic long to create the seed, it is possible for two threads to end up with the exact same seed. I could reproduce this close to 100% of the time when using 3 threads that all attempt to send at the same time.

Potential fix is here, although using putIfAbsent may be better:

https://github.com/stuartwdouglas/jboss-remoting/commit/14c6c7cb91700dd8340ec8f141b373f6203f77b5