We recommend that you always run a production system with some diagnostics switched on. This is necessary so that you can troubleshoot unexpected problems. Diagnostics consume resources, and there is always contention between problem determination facilities and performance. You need to decide how important it is to be able to diagnose a problem the first time it occurs rather than having to try to recreate it or having to just turn on diagnostics and wait for it to reoccur. We suggest the following level of diagnostics as a minimum:
You might also need to turn on tracing for task control (KCP), storage control (SCP) and application container (RTS) components.