I use metaphors like this when talking about programming: Any system of systems complex enough can easily be anthropomorphized...
But they both come back to basics: What is the overall model of how it should work? What are the symptoms? What do we know about the possible underlying causes of those symptoms? How do we design a test to see if the hypothesis is right? When we run the test, do we get confirmation, denial, or something else entirely? Did the test properly test the right thing?
The biggest part for me is the “reducing complexity” part. What are the intended bounds of operation for this piece of the system. What upstream effects can hit it or what downstream effects can it create.
A lot of analogies here with fuel systems in cars. When a car isn’t “running right” we generally start with the basics: is it getting air, fuel and spark, and go from there.
But they both come back to basics: What is the overall model of how it should work? What are the symptoms? What do we know about the possible underlying causes of those symptoms? How do we design a test to see if the hypothesis is right? When we run the test, do we get confirmation, denial, or something else entirely? Did the test properly test the right thing?
And so on...
And at some point, is it worth fixing?