On my regular job, at Canonical's MAAS team, we do, and we do a lot. The nature of the project makes it easier: it is a very specialized tool for managing primarily physical servers. Our users, therefore, tend to be very knowledgeable and have a very diverse set of problems they need solved.

That, of course, does not mean we don't say "no" or "not just yet".

Even on my own personal projects I tend to say "not yet" more often than not. Sometimes the feature requested makes sense, but the suggested implementation requires some thoughtful design. I've painted myself into corners far too many times before to fall into that kind of trap. As the Zen of Python says, "never" is often preferable to "right now".