It might mean you're not working at a high enough level of abstraction. If a task takes, say, 5 minutes, then it's pretty easy to get it done without distractio

It might mean you're not working at a high enough level of abstraction. If a task takes, say, 5 minutes, then it's pretty easy to get it done without distraction. If it takes considerably longer (and it's not very novel) then you're more likely to succumb. So, one thing you can do is try to develop better abstractions (either by learning to use or build better tools). In short, you want to put an upper bound on the amount of time you spend discovering and a lower bound on the amount of time you spend producing. So I suggest using your science dalliance to learn how to make your engineering efforts take less time. When the engineering time is suitably bounded, then you can use you broaden the scientific exploration a bit.