Sometimes we took a shortcut that hid the problem. Often the shortcut is unrealized at the time. (Using a lab power supply and not the battery the customer used. Opps.) Other times we didn’t think to test it that way. (Shallow water and slow moving boat. Either one ok, together, no good.) Then there is a whole class of low probably issues. Most units don’t show the problem or nobody is aware of it. Sometimes it becomes easy to see only when you get a flash of understanding that allows you to carefully design the right experiment. Only then it becomes painfully obvious. A wonderful problem is when a customer dreams up a new application and an opps pops up.
If you are lucky enough to have, what I call, Consenting-Adult-Customers, that that you can help test your product, it is wonderful. These are customers that you are comfortable showing your dirty, holey under ware to along with your products that don’t quite work the way they should. They will work with you to identify and fix your mistakes and not hold it against you and later become your advocate. They usually have experience that tells them, that, of course you are screwed up, but it is normal, human and ok. Treat these people very well.

At screening meetings I ask, “What are you going to do when you find out that it doesn’t work on Thursdays when the moon is full?” You will have a disaster, I guarantee it. I want to see if they thought about it and have a plan for it. In the end, it is really all management can do.