First, practice recognizing when you are bikeshedding. If you start from the assumption that bikeshedding is awful and evil and indicative of being a terrible person, that judgement is likely to disincentivize you from accurately recognizing it. ("only bad people bikeshed, and I'm not a bad person, therefore what I'm doing probably isn't bikeshedding" kind of thought flow).
FWIW, I've noticed a pattern in software bikeshedding where a team tends to bikeshed on the largest part of the issue that makes sense. In the classic example, "how to build a nuclear power plant" is too big an issue and people don't have the shared vocabulary/values to reason about it, or even to articulate what concepts they need to improve their shared vocabulary for. The actual power plant design is all uncertainty, with no foundations of shared understanding to build from. "Should there be a bikeshed?" is too small an issue; it's quite obvious that people will work at the facility, and might bike to work, and thus a bikeshed would be useful to have. It's all obvious (shared understanding) and no uncertainty. "How should we design the bikeshed?", though, is right on the edge: There's a foundation of shared understanding to work from ("it would be good to have a bikeshed"), and yet there's also some uncertainty/disagreement, because there's no one obviously right design for it. So stepping back and looking at exactly where a group falls down the rathole into bikeshedding is actually quite useful: it shows what the group already agrees on, what they have enough shared context to communicate about, and what they lack the shared context to even start discussing. To bikeshed the question, it seems like you're asking more about individual procrastination than the classic group bikeshedding dynamic :)
I find it helpful to view "bad" behaviors like procrastinating, bikeshedding, etc as carrying useful signals about one's internal state or the nature of the problem. For instance:
If the tangential idea seems more interesting than your research, why? Did you know that the research seemed less-interesting when you committed to it? You probably had good reasons to think that the delayed gratification of the research was ultimately more rewarding to you than the instant gratification of other things you could be doing; can you reconnect with those reasons? Or perhaps the research seemed much more interesting when you started the project. What changed -- does the current part of your work seem disconnected from the interesting parts? Perhaps you're catching onto some pattern in your work so far, which you haven't yet looked at closely enough to articulate, that suggests your current task is less likely to be useful than you'd assumed when you committed to it? And how are you certain that the interesting idea is truly unrelated to your research goal -- can you brainstorm a few ways that it could conceivably connect? Are any of these new ideas useful to your main task?
What's wrong with starting the essay? Are you afraid of failing, afraid of being judged by someone you respect if you don't execute it well enough? Are you worried that you haven't done your research well enough beforehand, or have missed something important? Does your outline for the essay look superficially ready but not feel right; do you need to go back and figure out how to connect the ideas in it together in a way that flows better? Does anything surprise you when you examine your resistance to starting?
Is your difficulty focusing telling you something about your internal state? Is your brain signaling "this is the best I can do on what you are/aren't feeding me", or "this is the best I can do on how much stress and how little sleep we're getting"? If at that time you have the cognitive capacity for mechanical tasks but not for creative ones (formatting rather than writing), is there a mechanical task you could do that would make a meaningful improvement to your essay (perhaps putting together all the citations you'll need, instead of worrying about the font size?)? What does this experience teach you about how you can schedule your writing tasks for times you'll be in a better state for writing?
In other words, I'm interested in ways I can design my work-flow/environment/habits to avoid bike-shedding (aka the Law of Triviality), which is the behavior of "substituting a hard and important problem for an easy and inconsequential one" [1]. Examples include 1) looking into an interesting idea that you ran across while doing a research task, even though it is irrelevant to your goal, 2) spending unnecessary time on the formatting of an essay rather than on the actual writing, 3) buying things/building systems to make very minor productivity improvements instead of doing your tasks for the day.