After upgrading JDK to the 8u191 release, CloverETL Server (versions 4.8.3 and older) starts crashing completely while running memory-intensive jobs. The Server GUI becomes unresponsive, soon the connection is lost, the whole application server goes down and needs to be restarted afterwards. This issue can potentially happen to newer versions of CloverDX/CloverETL Servers as well but it would be significantly less probable and would affect the Worker only (the Server Core would remain functional and the Worker would be restarted automatically).
It is related to the direct memory usage which is enabled by default in the CloverETL Server 4.8.3 and older. The issue itself appears to be determined by a bug in the Java 8u191 release.
The only "bulletproof" solution is to downgrade your Java to a pre-8u191 release until a new release is published correcting this bug. Another option is to disable the direct memory usage in the engine properties (this is described in our documentation for the CloverDX Designer engine properties but is essentially the same for the Server) to minimize the probability that this issue will be encountered.