Better… it shows more reasonable numbers.
I tried asking the metric from the master node directly (not via LB), it shows a number which seems right.
I have 5 nodes behind a load balancer, only if the request goes to master node, the metric value is from a sensible range of values.
ATM I have the In/Out values “crashing” from time to time, and the journal utilization & messages deleted from the journal warnings appear. A bit strange behavior, log messages continue to flow into the configured streams and are searchable. After re-starting the services (master started first) the in/out in the top of the webui starts showing ~15k/15k msg/s again.
Maybe I have to write a script as a dirty workaround to auto-restart-if-conditions-met while waiting for the next update.