Same issue with 16.3.6684.
At the exact same time of day (12:00), CPU increases to 50%, then after 1 hour it goes ballistic to 100%.
CPU usage chart from PRTG:

Again, all the services are stopped, there are no connections to server. 

Nikola, if you are willing, I would suggest opening a ticket support. We can then provide you with a profiling build and if used in conjunction with JetBrain's dot Trace and dot Memory at the time in which the server CPU increases, we will be able to gather very detailed dump information.

Was there any resolve on this issue? Even with the latest builds I'm having SmarterMail still eat up 100% of my cpu pretty constant and about once a week or so having to restart services to get webmail to work at any decent speed.

Adam, can you run a file proc on your server and see if any particular calendarData.xml file is getting written to more than usual?  I have had a couple of tickets where there were calendars with hundreds of duplicate entries which was directly contributing to the slow down.

I am seeing the same issue after upgrading from 16X to the more recent major build. No idea what triggers it yet but the CPU will go from average of 5-10% to all of a sudden pegged until I restart the service, around 24 hours later, the same thing happens again.

thanks Alex