1) First of all, make sure you understand the problem. There are no stupid questions. Do you understand what your client/boss is asking you versus what they need?
2) This will happen. "Build me a bridge by tomorrow". Make sure that you know for a fact that a problem is unsolvable within your constraints. Your client/boss might be flexible on the time/budget and these can be modified to give you more time/budget.
3) If the problem is understandable and the constraints are within reason, and there is technology that can solve the problem, but you just don't know enough...that's what StackOverflow and the Internet is for. Make sure you do your research first. Try to ask explicit questions that have quantifiable answers. Ask your peers. Have a design session.
4) This is a variant of answer number 2. It seems like your client/boss is asking the impossible. Do some research. Never say that the problem is unsolvable, unless you know exactly why and you can clarify.
5) ROI stands for Return On Investment. This refers to an investment in time. Your Time!. Is the problem important enough to solve to warrant the amount of time it will take you to research and solve the problem. Discuss this with your client/boss
6) Is it a real problem. Clients, often times, understand what they want, but don't necessarily understand what they need. Try to understand what your Client/Boss actually needs and discuss this with them.
Hope these guidelines help