Comment by ptaipale

I have a slightly more boring anecdote. We were making a 2 Mbit/s multiplexer (E1 to POTS and ISDN lines thing) and ISDN lines worked fine during the day in the test lab but they were always dead the next morning. The unit was otherwise running fine the following day but ISDN was stuck. We never got this to repeat at other times. If you booted it at midnight and came to work at 6AM, it was stuck. But it would run OK from 6AM to midnight so it wasn't any buffer overflow or such.

So, eventually, just decided to stay at lab for the night. At 5AM the air conditioning came on (it was shut for the night). The ISDN circuit got stuck.

It was electromagnetic interference from the massive AC unit fan which was not too far above our lab table. Improving the grounding of the ISDN chip helped.