As I mentioned in the support chat - the issue is that midnightPSTinUTC isn't dynamically calculated for midnight today -- it's midnight PST on Oct 4th. [i.e. over a week ago].
What's happening is the call is accepting that past date - which puts the script in the front of our queue to be run ('cause we think we're late!) - and it gets run in the next 60 second job interval. And then reschedules itself in the past again.
So basically this script is running roughly 60*24 = 1440 times a day.
Anyway - I'll have the devs look into it - we should return an error if you are scheduling the script too far into the past... [i.e. a second or two in the past is probably fine - and best to do in case the scheduling script has delays of some sort... but clearly 10 days in the past is too far! Probably 60 mins in the past is too far! 🙂 ]