For the record on server that don't run control panels, manually managed servers in a cluster like this:
2x haproxy load balancing servers.
2x nginx webservers.
3x database servers that are in a percona cluster.
and guess what, 1 of the percona servers gone mental due to the kernelcare automatic update. From that day the decision is to NOT automatic update kernelcare.
Now it's done once a month with an engineer that keeps an eye on it.
IT is lovely! 😄