Summary: Build a PROPER reader centered structured (not freeform) reporting tools, and DON'T use templates, wikitext or jargon.
I'd say the problem of reporting issues and uploads is not really with the newbies or editors. It is an issue of centralization and commons power users / admins.
The issue here is that both WMF and Commons admins think too small. Wikimedia has millions of readers. If every single one of them was exposed to commons images and asked to evaluate it for usefulness as "educational material" or general usefulness, this would make it possible to quickly provide useful meta-information about images. You'd quickly create a database of image evaluations that could potentially surface the really bad images. The important thing is to not allow free-form data, as this would need curation and moderation. Also randomly exposing these images to editors from every wiki would decentralize the curation and evaluation of commons only users, and allow more eyes and ears on the content. While most people have limited understanding of copyright, everyone can spot an obviously bad quality image.
Instead of text only captcha, you could have a sort of mediacaptcha, that surfaces these images / media, and asks questions like "is image a selfie", "is image an object", "is image educational", "is image inappropriate", "is this image a screenshot of a movie (copyright)", and so on, along with the normal captcha. They'd only need to solve the normal one correctly to continue editing, and the information from the mediacaptcha would be aggregated, and used to help improve articles. This would probably need to use some check for age, as commons contains lots of culturally inappropriate trash from time to time.
I think the best way forward for both media and content in pages is the tool described in this : https://www.mediawiki.org/wiki/Requests_for_comment/Tickets
As a reader, I'd currently never use a talk page to report anything because it is a really bad tool that has been abused for things it was never designed for.
Page triage is a good example of a reasonable tool that ended up being inflexible due to its heavy dependence on templates, and wikitext.