Very late to the party here, but I thought adding to the existing discussion would be more useful than starting a new thread.
In a compromise with the one-big-database and different-databases-for-different-topics positions, I use two databases: One for my main research area (Foreign and Second Language Writing) and one for everything else. (A few items that intersect multiple research areas are held in both databases.) I know that it would be possible with skillful use of keywords and groups to replicate the Writing database view within one comprehensive database, but it seems easier to me to know that everything to do with writing (and nothing else) will be in my Writing database, and it means that I don't have to add the keyword writing to thousands of items (or put those thousands of items into a static group). It also means that, with my current iPad running out of space, I can sync only the Writing database if necessary. In addition, if one database became corrupt and I had neglected to create a recent backup (it did happen to me once), I don't lose everything. (I know this isn't great reasoning; if database corruption was a primary consideration, I guess the different-databases-for-different-topics strategy, coupled with a very rigourous backup plan, would make more sense.)
In the past I have tended to favour smart groups, based mainly on manually-assigned keywords. In order to avoid confusion between "sciency" or actual content-related keywords becoming confused with organizational ones such as those referring to the papers in which a reference item is cited, I have prepended the latter type with a special character. I used to use @ as the special character in all cases, but I now tend to use that with the names of students or friends to whom I want to recommend a paper and prefer $ for prepending to the names of papers in which a reference item is cited.
I should note that I haven't conducted or referred to any studies into whether it's quicker to drag a paper to a group or to type a keyword, but I've had a general feeling that it might be advantageous to have important information like that actually contained within a field belonging to that reference item rather than being part of the whole-database information that certain items belong to certain static groups. (If this is an illusory distinction, please disabuse me thereof!) I think I've also unconsciously carried over habits from other applications that don't allow for much hierarchy and that thereby put lefthand pane real estate at a premium. Because Bookends has folders in addition to static and smart groups, this consideration isn't really relevant.
Recently, though, considering convenience, I've been thinking that static groups might be easier. First of all, it's easier to drag items between groups (or change the name of a group) than it is to change tags in multiple references (such as changing @ to $). Additionally, smart groups don't sync over to the iPad, whereas static groups do. Keywords also don't really do anything on the iPad; as far as I can tell you can't tap on a keyword and thereby see all the items with that keyword. Although I'm happy to add items and do all my organizing on the Mac, I do like to be able to leverage that organization fully for reading and browsing when I only have the iPad with me.
As for integration,
(a) For writing, like many others, I use Scrivener 3. When I've finished my first draft of a paper, I sometimes use the basic function of selecting the cited references in Bookends and generating a formatted bibliography, which I copy and paste into the paper. I keep track of which papers to include "manually", using the methods mentioned above, and don't use any cite-while-you-write features. In my draft in Scrivener, I tend to insert the citations manually (Crawford, 2003); or put the full formatted citation into the Notes pane or in the main text right under the place where I mentioned it; or, most often, I create a new text in the same folder and show it in the righthand pane of a split window, and add formatted references to that as I cite them. You will understand from this that I am quite happy after finishing the content of the paper to fiddle around tidying up the text, arranging references in alphabetical order, etc. Maybe this is misguided, but it seems like such a small task, even if there are 100 or more references, compared to actually writing anything any good.
(b) For leveraging my Bookends database to keep track of information, generate ideas, etc., I like to put notes from Bookends into DevonThink. If I'm working with a paper that has a good literature review section I tend to take copious notes in Bookends using Command-Control-P to capture the exact wording used by the author along with the page number. If it's an important paper, I also take the time to add a title to each notecard as well as %tags to some of them. (As far as I can tell, those tags are not as useful as they could be: it isn't possible to generate tag clouds or browse using them, but it is possible to search for them, which in itself is very useful.)
What follows again applies only to important papers, but, once I've taken copious notes in Bookends, I export all the notes into a folder in DevonThink Pro named with a shortened version of the paper's title. I add a full formatted reference to the Comments field both of the folder and of every note therein. In DevonThink, in addition to folders dedicated to one paper, I also have a top-level folder containing a whole load of Subject folders. Selected notes from specific papers are replicated to the relevant Subject folder(s). For example, in a paper on vocabulary gains to be expected from certain amounts of reading, I read that:
Milton and Meara (1995) tallied learners’ vocabulary growth in periods abroad not at 1,000 but at 550 words a year.
That's useful information for my Study Abroad folder so I replicate to that folder. The entire note looks like this:
{Cobb, 2016, #16734@301}
@301 Milton and Meara (1995) tallied learners’ vocabulary growth in periods abroad not at 1,000 but at 550 words a year.
Milton, J., & Meara, P. (1995). How periods abroad affect vocabulary growth in a foreign language. ITL Review of Applied Linguistics, 107/108, 17–34.
Cobb, T. (2016). Numbers or Numerology? A Response to Nation (2014) and McQuillan (2016). Reading in a Foreign Language, 28(2), 299-304. Retrieved from
https://files.eric.ed.gov/fulltext/EJ1117024.pdf
bookends://sonnysoftware.com/16734
This is where things are a bit messy and haphazard in terms of workflow. I think I have everything I need there: the actual quote, where it comes from (the Cobb 2016 paper), the orginal source of the information (Milton & Meara 1995), and a quick way to get from here to that information in Bookends. But I often forget keyboard shortcuts and still am uncertain what the fastest way to accomplish this is and tend to include redundant information.
Apologies for the ultra-long message. About 16 months have elapsed since the older messages in the thread. I'd be very interested in hearing about any evolution in people's thinking and workflows.