I'm currently preparing a manuscript for publication, and will likely upload it to the arXiv first. However, let's say I'm submitting to a journal with a very identifiable formatting style, like, let's hypothetically say, an IEEE journal. While most articles I see on the arXiv are fairly "format agnostic," I wonder: is it frowned upon to upload a preprint using the aforementioned distinct formatting style (provided that all mention of the journal is excluded from the preprint, of course)? Little clarification is given on that specific question by the journal's guidelines. Should I instead downgrade the article to a more bare-bones style/format?
No, it is not generally frowned upon. Using the same format as for the journal submission is less work, and commonly done regardless of what information content it might provide. In my field I often have a decent idea of which target journal a given preprint is written for. And sometimes I see formatting changes in the next version indicating a lower-prestige journal. Some authors might want to make the submission history less transparent by using a more generic format, others don't worry about it. As far as ArXiV's instructions are concerned, all they say is not to include actual journal template files in your submission, but any documentclass they support is fine.
[The following isn't strictly an answer to the question, but rather an argument for why the question doesn't need to be asked.]
In my area (pure maths), it was a bit of a revelation to me to realise that although most journals have quite specific formatting guidelines, it does not remotely matter whether you follow them in your initial submission.
The process of getting your paper refereed, and deciding whether or not to accept it, is in the hands of academics -- the editorial board and the referees -- who couldn't care less about the publisher's house style guidelines. They aren't going to reject your paper because the margin is 2mm wider than the guidelines, or the font is Computer Modern instead of Times; they will only even notice the formatting if it gets in the way of reading the content [*]. The only people who care about formatting niceties are the journal production staff working for the publisher, and they will only get their hands on papers after they've been accepted.
So there is no point (in my field at least) in spending ages reformatting your paper for the guidelines of the specific journal you're submitting it to. Just choose a standard generic latex template like
scrartcl or whatever, and use this for everything (preprints on your personal webpage, preprints on the arxiv, and submissions to all journals). You can worry about reformatting things to match the journal style later on, if the paper gets accepted -- which means you only have to do it once for each paper, even if (as usually happens) you need to try multiple journals before one of them accepts.
[*] There is an exception here: one factor involved in accept/reject decisions is the length of the paper, and if you are obviously trying to game this by using really narrow margins and tiny font sizes to make the paper look shorter, then it will annoy the referees and editors. Evidently, annoying these people is not in your interest. So using latex packages like
savetrees is probably an own-goal -- these are sometimes useful for grant funding applications, where the page limits are strictly enforced, but don't use them for journal submissions.