@mntmn I once had an ARM LPC2106 dev board with crosstalk from the RS233 lines onto the processor reset line, that under high speed RS232 traffic would inject short negative pulses into /RESET, locking the CPU up as they were far shorter than the datasheet mininum reset pulse.
You can imagine the fun I had working out this was the problem, not a bug in the UART driver I was writing 🤬