Questions on URL sources versus text sources

I've been trying to fill in image sources for some past uploads (mostly over a few years old), but in some situations I'm not sure what's the best kind of source to use.

I've considered the help:image source wiki page, particularly this part on non-internet sources:

Non-internet sources

Note that providing a URL for your uploads should be preferred over all else, especially if it did not come from yourself. However, for uploads from a non-internet source (self scan, fandisc, screencap, game rip, etc.), specify what material the image came from: the work title or the episode name, for example.

1. Suppose we have a scan of an artbook page, and it's an md5 match with a third-party post (not posted by the artist) like yande.re. Example: post #590525. What kind of source is preferred? A yande.re URL, or source text with [Circle (Author)] Title of the artbook? Note that the yande.re post might be part of a pool which identifies the book clearly, or the yande.re post's source field might be filled in - would that affect the answer?

2. Suppose there's an image where someone added a third-party URL source many years ago, but now it's a broken link. In the absence of an alternate URL, is it better to replace that source URL with a text source such as [Circle (Author)] Title? At the moment I'm mainly thinking of moeren and Dropbox sources (source:http://moepic, and source:http://dl.dropbox plus the https variant).

3. Suppose we have a scan of a dakimakura, stick poster, etc. and there's no known md5 match anywhere. Often, it may be impossible to find a word-for-word official name for this kind of product. In this case, which source type is preferred: an informational URL (like a product page or blog post) which identifies the product well, or a text source describing the product (including brand/vendor, release date/event, etc.)? Are informational URLs acceptable as sources, or should such URLs go in comments only - say, due to non-web source tag gardening concerns? Take post #1475845 for example.

Apologies if I missed some obvious answers. I searched around the forum somewhat, but didn't quite find what I was looking for.

Well, I did some more reading, and so far I'm thinking this:

The top priority I'll aim for is to preserve and provide as much relevant source info as possible. That means both a URL and text if I think just one might be a bit inadequate. There are multiple places where this info could go. I can't find a catch-all consensus on which info should be presented where - so if someone else has indicated a preference (such as by filling in the source field) then I won't override that, but if it's left to me, then I'll use my discretion.

Here's the source info we might want:

• Info on the artist, and on the larger work that the image is a part of (book, etc.), if any.
• Including informational URLs like those from doujinshi.org (topic #11093).
• Info on where the exact image file originated from (forum #146933, forum #124444).
• All else being equal, prefer sources that are more accessible to the public, such as yandere over exhentai (forum #152714).
• Even if I don't know what site the uploader actually pulled the file from, listing an md5 match from a site like yandere could still be informative.
• Regarding moeren, there might be archives of image boards; perhaps a downed URL can be replaced with an archived version in that case (forum #124421).
• Regarding Dropbox source URLs, Danbooru has or had a 10 MB limit for direct file uploads. So a workaround is/was to upload to somewhere like Dropbox first, then do a URL-based upload from there to Danbooru (topic #8584). This probably explains most of the Dropbox source URLs. Still, being able to imply "this is most likely a self-upload" is still SOME kind of sourcing info, and maybe not worth replacing with a text source.

And here's where we might put that source info:

• In the source field.
• This field may or may not be subject to gardening concerns - for example, md5 mismatches, or scripts related to the non-web source tag (topic #15647).
• In a comment (forum #146772).
• Comments have more flexibility than source fields: they can have multiple informational URLs with descriptions of each, and have room to provide both Japanese and romanized names.
• However, the source comment might get ignored/buried if there are several other comments on the post.
• The vast majority of users can't edit other people's comments, so any corrections/updates would have to go in another comment.
• In a pool title and description (topic #11093, forum #124444).
• Making a pool for a lone book cover - or in general, a clearly incomplete set of items - might overly encourage others to upload the rest of the book (unauthorized). I'm not sure about preferred countermeasures, but I could potentially add this to the pool description: "This pool was primarily created to source some old posts. Make sure you know the site rules before uploading more from this work."
• I would probably use this option if my alternative was to write substantially the same comment on several different posts. For example, if an artbook has several pages uploaded.
• I didn't find anyone stating the idea of "if a yandere post is linked and that's part of a pool, then Danbooru doesn't need to repeat the pool's info". So I probably won't follow that idea.
• In a tag's wiki entry.
• Game CG is probably the closest example where almost everything that would go in a text source can go in the wiki. Though, a lot of prominent games have a fandisk or append patch, and it's still useful to note whether a specific CG is from the base game or the fandisk/patch.

I think I'm mostly satisfied with my search at this point, but please do point out if any of this sounds a bit off.