In looking at the code, we definitely pull over *all faults* sorted, then pretty much throw out all but the last ones in almost every situation.
I expect what we need to do is create an optimized query that only returns the latest faults for each uuid. That would dramatically eliminate content on the wire for long lived instances that might have seen issues over time. That needs more clever SQL than my brain can muster right now. Will engage Jay Pipes in that.
In looking at the code, we definitely pull over *all faults* sorted, then pretty much throw out all but the last ones in almost every situation.
I expect what we need to do is create an optimized query that only returns the latest faults for each uuid. That would dramatically eliminate content on the wire for long lived instances that might have seen issues over time. That needs more clever SQL than my brain can muster right now. Will engage Jay Pipes in that.