Hi Pedro,
@andreas-doehler said it pretty well, there is not one single answer. Many thresholds are also based on experience that our customers and partners and we ourselves have made over the years.
In many cases there isn’t even one “true” answer - depending on the use case, the criticality of the monitored host or service, the volatility of the performance metrics you may be OK with a given value, while another customer is already pressing the panic button.
A CPU load of 98% may be fine for a test system, but do you really want that in production? Etc…
But is there any type of devices (a particular vendor, for example) where you have the impression that the default thresholds are particularly off?
Best
Elias