Hmmm... I'm currently a bit short on time, but if there is one thing that I can assure you with 99.9% of confidence: it wasn't this hack.
There is some odd behavior of the device which I need to take a closer look at (as I noted above: /mnt/us doesn't even (necessarily) seem to be mounted when the framework starts), but this entire thing is as failsafe as I could make it. I took all the precautions that I could to prevent this from crashing, let alone permanently bricking the device. If something goes wrong, this hack detects it and stops mucking around before anything serious happens. (*)
That said, anybody else willing to give it a try? (It may not become active on the first run for the abovementioned reason, but restarting the framework manually ["restart framework" or "killall cvm"]) will trigger it.
(*) That said, there is always that 0.1% chance which I'm not categorically excluding. If this hack really screws up things, I'll be delighted to get some logs / stack traces.