I wish someone with a (fair) bit of Linux knowledge would compare the USB traffic when running the serial-loop sketch that locks up the board.
It would be nice to see the difference between an UNO and one of the FTDI-based boards (Nano, Dicemilia, ...).
I've tried this (using wireshark), it gives a thoroughly unpleasant mess of message packets going back and forth. Unfortunately comparing a UNO and a Duemilanove isn't helpful, because UNO uses a standard CDC-ACM protocol but the FTDI uses its own protocol and they are not comparable. What you need is another device using CDC-ACM - some mobile phones do.
I agree that they really should give linux users a warning that the software isn't quite ready yet. The best advice from my own experience and from the boards is this:
1) Avoid using serial if possible
2) If you have to use serial, put delay(5000); before Serial.begin to give yourself a chance to open the serial monitor or gtkterm before the output starts
3) Avoid the tools menu as much as possible and never select speeds of 14400 or 28800 in the serial monitor
4) Don't unplug the Arduino and immediately plug it back in. Wait for a few seconds before plugging it back in. Otherwise it can change ports from ttyACM0 to ttyACM1, and then you will need to go to the tools menu, which you don't want to do.
5) If you are really stuck and cannot upload to the device, unplug the Arduino, wait a few seconds, then hold down the reset button for a few seconds while plugging it back in. Then upload.
Has anybody else managed to get a working LUFA setup on linux for updating the firmware? I've been experimenting with altering the firmware and had some success, I wondered if anyone else had tried.