I work in academia (research). Tasks are varied and almost always involve delving into topics that produce more tasks when you are more informed. A pull system seems appropriate rather than an estimated timed-box scrum.
My question is then - is time-logging really recommended? Projects are so varied, that I would not be comfortable looking at a few previous projects and saying "this new project would take the same as these previous projects". The goal for us is to be as lean and flexible as possible to have a high-level visualization of each project to know where we are with each at a glance. If new projects come in, we'll have an hour meeting to discuss it and use our experience to give an overall estimate of when we can deliver the project.