It may be useful to spend some time with a pencil and paper and tutorials like this one: (I googled "tutorial cartoon animals")
https://design.tutsplus.com/articles/cartoon-fundamentals-the-secrets-in-drawing-animals--vector-17361
Even if you don't find one for exactly the drawing style you prefer, the practice is really helpful to represent what you want to draw - and understand what makes a character look happy, grumpy, etc.

Painting over a scan / sketch can be very in-depth - but in the example that you show, it's rather clear that the buildings retain a lot of pixel details from scanning, with as little re-touching as was necessary.

it make me wonder why people are criticized if they use these methods nowadays to make pixel art.

Mostly the issue of plagiarism (if the source is an artwork) and undeserved praise.
It's not so new, because even 20 years ago there was the "no copy" movement in demoscene, as a reaction to the (over?)use of copying paintings, with and without credit. Boris Vallejo, Luis Royo, and licensed AD&D artists got copied a lot.
Nowadays, galleries like Artcity (artcity.bitfellas.org) and Pixeljoint  strongly advise providing sources and references, so that the viewer can make their own opinion.
Making-of videos can be very interesting to watch (or at least skip through), it's not unusual to see that the video is at 33% duration when you'd consider the image finished, but a dedicated pixel artist will keep tuning and fixing and redoing parts that he doesn't like.

Unfortunately, I feel this is killing the site quicker than a hostile takeover. There is everything from straight "wall of text" spam, to hidden links in copies of old posts, and con artists posing as newbies in order to beg for answers.

AFAIK, it's well known. You need considerable experience with a topic in order to draw / improvise from your own memory, so photographic material has often been used as reference (so you can frequently watch what you're trying to represent), tracing material (when you copy the exact silhouette / proportions), or straight up pixel base (you paint over an already digitized copy).
Note that this is never a guarantee of saving time : converting a digitized photograph to your current style and environnement (lighting, colors, resolution, level of detail) can be more time-consuming than straight up drawing, and only very skilled pixellers can guarantee that they reach the style of the pixel graphics around.
I think it's like special effects in cinema : any time it's hard to judge if something is real or not, it means somebody did a good job.