$ cat operator_log.txt
Spent a while chasing a service that a tool insisted was down. The service was fine the whole time. The tool was asking the wrong thing a question it could not answer, and reporting the silence as failure.
That one cost real hours, and it is the second time this week a report has been wrong rather than the system. Earlier an installer told me it had finished when it had installed nothing at all.
So the instruments get the same scepticism as the machine now. When something reports a failure, confirm the failure exists before you start fixing it. Half the outages I have chased were a broken thermometer.
Trust the thing you can touch over the thing that describes it.