It was down as of 7:45am (we posted in our engineering channel), so that's a good 40 minutes of public errors before the root cause was figured out.