In a startup I've worked on, we did estimates based on hours, but also had the "no idea" tasks. We would then make sure to alternate sprints to work on well est

In a startup I've worked on, we did estimates based on hours, but also had the "no idea" tasks. We would then make sure to alternate sprints to work on well estimated tasks (as you say, "linear" tasks) and others to work on the hard ones, without a set deadline; or split the team effort in that manner. I think that's a better strategy than making wild guesses and ultimately falling behind schedule, but at the same time maintaining cadence, which buys you power to sometimes say "hard things are hard, I don't know when it'll be ready to ship".