When I'm addressing bugs in the project I'm working on, especially when the bugs are described vaguely by outsiders, I start by confirming what I know to work. Take the most likely path I can conceptualize them reaching some sort of issue that's similar to their reported issue. While confirming what I know to work, if I'm not encountering anything that indicates that I'm on the right path, I switch over instead to what potential edge cases could be the issue. Focusing on the expectations, assumptions, and behavior I expect to see given what the options are provided (do they have the option to switch between two screens over and over).