@niconiconi Crash it due to not upholding expectations around rise/fall rates (and something in hardware starting to have problems in the direction of metastability) or because of software that deals poorly with that (by e.g. unmasking the interrupt too early or sth. like that)?