I've also been noticing this issue - and it seems to have started at midnight here as well. I can reproduce it on all test systems, none of which has shown this issue before.

I'm still investigating this right now, but early analysis suggests that computation of a date for a timer goes wrong due to the daylight saving time. The timer then fires all the time (instead of once a day), causing 100% CPU load.

So chances are the issue will be gone in a few hours.

Best regards, 
Felix Schwarz