Hi,
  I've experienced the same problem, two nodes got stuck with CPU at 99% and the following source code from IncomingStreamReader class:

while (bytesRead < pendingFile.getExpectedBytes()) {
                bytesRead += fc.transferFrom(socketChannel, bytesRead, FileStreamTask.CHUNK_SIZE);
                pendingFile.update(bytesRead);
            }

the cycle seems to loop forever...
after both Cassandra instances were restarted the problem didn't occur. 

Best regards, Daniel.