Currently, the average CPU usage on SCB nodes is about 60% which is too high for normal operation. If we loose one of the boxes we would be in trouble since a single box can't handle the load any more. Today it was forgotten to repool scb1001 for an extended period of time and Mobile-Content-Service was experiencing problems (worker deaths, monitoring flapping, occasional timeouts and 500 errors).
After transclusions processing was enabled in ChangeProp request rate to Mobile-Content-Service can grow significantly when some highly used template that changes html is edited, so the request rate could get as high as 100req/s. That obviously increases the load generated by MCS. As the amount or processing done by MSC grows, this issue is becoming more an ore pressing.
So, we either need to prioritise T96017 or get at least one more box for SCB