The Importance of Mental Metrics in Data Science

In any distributed system, which is where SREs are gonna be, there's just so much I can go wrong. It's almost like things don't break in a very binary way too. So you mental metrics aren't great or maybe are imperfect. Is there such a thing as like kind of a set of metrics for an ML platform team? You know, if you had to say one or do you think that in general was it the pandemic?