Drupal is a web CMS that empowers people without extensive coding experience to do some pretty amazing things. A front end developer at a small consulting company can build a useful experience pretty easily.
And if you are a developer you can extend it extensively.
It hit its peak before some of the extensibility features hit WordPress.
When it moved to being symfony based with Drupal 8 the shift was so great for this audience that Drupal hit its peak usage [1]. It's been in decline since then. Drupal 7 is still the mostly widely used single version and BackdropCMS came long to support that crowd (a Drupal 7 fork).
I look at this as an exercise in knowing your target audiences.
It caused much of the internal of Drupal to be re-written. This included how it was extended. With previous major versions you learned about new features and APIs. They followed mostly existing design patterns so it was easy to learn and updates your extensions for. With Symfony you had to learn whole new systems and ways of doing things. It was like learning something entirely new. And, porting extensions to it was far more work and time.
Also, the updates made Drupal slower while consuming far more system resources for the same thing. This increased costs to operate.
All that flexibility resulted in pages with 1k+ queries on them, and nasty upgrade paths. I hear D8 is Symfony-based now, and a lot got cleaned up, but after one project upgrading a Drupal app from one major version to another I won't touch it again. After a month's worth of work we refunded the client and canceled the job.
Last time I looked at Drupal (a few years ago) the majority of Drupal 7 modules had not yet been ported to v8 which had been released for some time at that point.
Wrote something called the "page module" that let them build out the various landing pages using extremely flexible criteria ... everything from tags to specofic articles to category - and probably more that I'm forgetting. I built specically because views just wasn't good enough. It was able to do a couple of very important things... It had what was essentially a materialized view that was maintained by the import scripts (All actual content editing was done in the propritary publishing software they used, then sent as XML to the web importer which ran constantly looking for new xml files in a directory).
The big upshot of this was that is that getting, saying, the articles for a given "page" (Which wasn't necessarily actually a page - there were also presentations like sidebar lists) was a SINGLE query with no joins. Very performant. We did millions of page views daily on one web server and one db server... both low end commodity boxes.
It was kinda fun, but when I left I made a promise to myself to never, ever, work with Drupal or PHP every again professionally, and I've been quite happy to stick to that.
I guess I wouldn’t say quite never on Drupal, but it would need to be like a 2x comp move to even consider.
Then, of course, once adopted - like most CMS - it's very hard to migrate out, and it snowballs.
People largely went for Drupal over WP or Plone or Django because they had zero leverage about 1) what comes in to the content system, 2) how they can build a content system, and 3) what comes out of the system. Also, in the aughts, don't forget that WP was extremely primitive compared to today. Lots of enterprise leadership said, "oh, you can't have Python, that's for engineers" combined with "you have to read and write all articles to this arcane XML schema only parseable to five academics in France". Thus the Drupal extensions ecosystem ended up with a TON of low level plugins . . like Drush, which probably shouldn't have existed.
EDIT My God, Drush is still around.
Speaking more broadly, Drupal was also seen less as a "framework" and more of a straight "CMS", so the idea of a CMS needing a CLI was, then, a bit of a "Huh?". CMSs were regarded, at the time, as the thing you set up so the secretaries and executive assistants could update the website. Of course, in the decade and a half since then, we've seen how CMSs inevitably grow into pseudo-frameworks.
We also have a better grasp, today, on just how complicated the functionality of a CMS actually is, laying as it does across the intersection of natural and formal languages. A linguistics specialist could have predicted this, or even a dedicated data scientist, but the content world is shockingly bereft of linguistic specialists among its spec writers and practitioners. DITA - to abuse one single example - arose almost entirely from pedagogical academics, with no input from linguistics, computational or otherwise. And DITA is actually kinda-sorta planned. The various MIL-STD content specifications - 48 and counting, not including the bespoke USN things - don't arise from anything, they're just eternal flypaper for whatever contract items get thrown at them, plus golden gimmes from the blessed vendors.
The challenge, I think, is that by that point in time Drupal had gone through its first big popularity explosion, and was starting to grapple with the competing interests of many different audiences. Acquia ended up being instrumental in steering it towards "enterprise sites with complicated UGC requirements," but for quite some time the open source ideal of "the project reflects the priorities of the people who contribute time to it" meant it lacked a strong, opinionated take on many of the things you mention.
For many years Drupal's strength was (IMO) that enough of _those folks_ existed in the community to ensure you could build complex, highly adaptable structured content systems with it... and enough of _those other folks_ existed in the community to ensure that there were click-together content display and delivery tools that worked with the complex content. If you approached it with a clear understanding of where some of those boundaries were, you could build really amazing things — but if you came in looking for a well-paved path to build a simple site or architect a complex one, well... ¯\_(ツ)_/¯
edit -- clicked on your profile and read some of your posts in other threads, and I feel like I should just get a beer and commiserate with you about doomed CCS initiatives for a few hours. I salute you.
As someone forced to use it and learn it for 2 years, I did what they asked me to, cleaned it up, and then convinced everyone to move away to a different system (detailed here: https://news.ycombinator.com/item?id=38489357).
We ended up with a system that was much much faster, much easier to edit, easier to maintain/develop, and thousands of dollars cheaper (compared to Acquia). It's the kind of thing that management would never have bothered to investigate without internal pressure (because, again, Drupal checked all the boxes). It's just an implementation detail that bigger businesses are happy to spend resources on, but it sucks for the day to day users.
It falls squarely in "never EVER again" territory for me too, like others here.
The onboarding experience is terrible, too. Compare Wordpress's editing landing page (https://wordpress.com/website-builder/) and editor documentation (https://wordpress.org/documentation/article/wordpress-block-...)
versus Drupal's content editing page (https://www.drupal.org/features/content-authoring), which says nothing at all and just points you to Drupal contractors and sponsors, or its documentation (https://www.drupal.org/docs/user_guide/en/index.html) which is super verbose but has no clear way of getting started.
The whole thing has that "made by developers, for developers" feel, and editors are a complete afterthought. Maybe that was the norm in the 90s and 2000s, but these days there are much more user-friendly, editor-forward options that give them a nice experience out of the box. Drupal never entered that era, and culturally it's still way more focused on the underlying technology than the end-user UX.
It's a CMS that serves admins & management first, developers second, viewers third, and editors last.
Drupal: Never again™