Hi @mattwarkentin !
I work on the biological science area and I ran into the same questions. Unfortunately, since I collaborate with people that won't migrate into reproducible documents workflow; I almost had to abandon this quest. And like you said, even some journals accept only msword files. Regardless, I gave it a long thought on how to try to adopt this practice, eventually rendering to msword for publication/collaboration. Assuming you are/will be using RStudio, here are my thoughts:
This is the most annoying part. There is no simple way to approach collaboration if people aren't willing to migrate to a markdown format. The best approximation I could find is the Trackdown package by Emily Kothe et al. It is far from ideal, and it requires everyone to agree on some workflow ground-rules, but it tackles one of the most difficult issues on collaboration (i.e: Tracking changes to manuscripts). It requires working on google docs files, though. I do not know if there are programs/platforms that let you achieve this in a seamless way. You could try Jupiter notebooks, but I find that the tex formatting is less appealing that markdown (specially if you are collaborating with non-tech savvy people).
Edit: Googling stuff for this post I ran into Manuscript. I haven't tested it, but maybe it's worth looking into.
Regarding rendering, I only render into msword, using a template docx file. Journal guidelines usually only require minimum page and typography formatting so you can easily create your own templates; if you can't find any online. Regarding Tex/PDF, if you are lucky and your journal accepts tex files, they usually have their own template. Other than that, journals will do their own formatting, so I fail to find a reason to try and do it myself.
Two ways. If you use R for data processing, or otherwise manage to load your data tables into R (from excel save as .csv and load it into R), I suggest you look into the kableExtra package by Hao Zhu. If you don't, I would stick with MS Excel/Word (or googles alternatives) for simplicity. I don't think markdown was conceived for complex tables.
Not an easy answer. Journal guidelines mostly focus on how to process figures and tables, how to cite, word count, symbols and abbreviations, etc. Most of these you will type, so read the guidelines and try to keep them in mind when writing. What can be standardized is citation style, font type/size, and page format. As I said, if you're lucky, the journal you want to publish with will have a Tex template, otherwise create a standard msword template that's easy to collaborate with (double spaced, numbered lines, etc.). Again, final formatting will be done by the journal. For citations, I recommend you use Mendeley/Zotero (or whichever reference manager you like) and export you citations in .bib format (make sure you assign a Citation Key to each reference) and cite in your manuscript as you write. Next, download the appropriate .CSL file (in github citation-style-language/styles or zotero styles) for your journal and set it in your YAML (see in 'R for Data Science' online book the 'YAML Header' section).
In conclusion, it is not a streamlined process. But that doesn't mean it isn't worth it. I'm currently using quarto/RMarkdown for writing my thesis. I created a RStudio project, a global _quarto.yml config file, and a .qmd file for each chapter/section. I plan on creating all my figures and tables with ggplot and kableExtra, and embedding them inline with the text. I have a draft template and will create a final version template for final proofreading and publication purposes. Drafts will be send to my adviser and I will use msword track changes function to manually pour them into the .qmd (this will be very annoying, but googledocs and trackdown is sadly not an option). The final version will eventually be converted into PDF. References are managed with Mendeley and an appropriate CSL. Lastly, everything is tracked with git, so I can easily revert any change, while also having another online backup (besides my online drive).
It definitely is too much if you simple want to write a manuscript (msWord with Mendeley's plugin work great for reference managing), but I hate the idea of having multiple chapter 'x' versions files; and since I already use R for figures and tables it isn't too big a deal for me.
Hope this helps anyone. It certainly took some time to write 
.
PS: Unfortunately I could only include 2 links -.-.
Cheers,
Tom.