There are definitely some outright errors in this article. For one thing, D&D 3.0/3.5 had plenty of art, though that's no longer obvious from a Web standpoint now that WotC's D&D 3.5 Archive is down and the various books' Art Gallery articles with it.

For another, it buys into the usual horrible, horrible misinterpretation of the Ivory Tower Game Design article. The idea wasn't that they intentionally seeded the game with trap options to screw over dumb players, it was that some options had specialized situational uses that they made no effort at all to separate out, indicate, or provide usage guidance for, instead assuming that it was OK to let people fumble through and feel smart for figuring those details out for themselves. It may well be no coincidence that 3.0's, d20m's, and 3.5's life cycles predate the codification of the Three Clue Rule. :p Indeed, Justin Alexander has an article about this very misinterpretation: https://thealexandrian.net/wordpress/2498/roleplaying-games/thought-of-the-day-ivory-tower-design

This is not to say there isn't plenty of valid critique of 3.5 in this article, though, and plenty more to say about it. I may love the 3.x editions (3.0, 3.5, d20m, and PF1), and think WotC's been driving the game in a different direction than I'm interested in since, but by no means do I think 3.x is perfect.

It is worth noting that the design philosophy at WotC at the time was that splatbooks should mostly ignore the existence of other splatbooks, aside from mission-critical dependencies like having psionic material require the Expanded Psionics Handbook, Eberron supplements require the Eberron Campaign Setting, and so on. Presumably the intent was to try to cut down on the degree to which books pointed at other books and made people feel cheated for buying incomplete rules. But the reality was that the devs were failing to take interactions between supplements into account... when the playerbase was newly empowered to find interactions by comparing notes and then share those interactions (as you so clearly spell out in your Forum Culture section). It didn't help that the devs were so confident that D&D 3.0 was perfectly balanced that in Dragon Magazine they essentially double-dog-dared the community to break it.

On the spell availability issue, there's another factor that made the 3.x Cleric extra-powerful compared to past iterations: their going from having to prepare almost all healing spells to their not having to prepare healing spells at all (not just spontaneous curing, but Wands of Cure Light Wounds). A lot of the powerful core spells for clerics already existed in the TSR editions, but they were impractical to use because the constant need for heals drove their prep decisions. Once cleric players could actually cast non-healing spells sometimes, they discovered just how powerful they really were... and then, if they were sufficiently self-centered, abused that fact to stop being a team player entirely.

While it's easy to blame supplements for the brokenness of 3.x even while acknowledging how powerful core spells were (and annoyingly, some people refused to do the latter; the author of the article has a leg up there), the seeds of Persistent Spell existed in the core DMG in the form of the magic item pricing guidelines being every bit as agnostic to the interaction between spell durations and duration overrides as Persist is. That was a massive design flaw, and one that most don't really give enough credit. If I could freely take all of 3.0, 3.5, PF1, and d20 Modern (and Dreamscarred Press material while I'm theoretically at it) and start with that massive library to make my own idea of what something with similar design sensibilities "should" be like, fixing that would be one of the more fundamental "meta-rule" changes I'd make.

Concerning people ignoring the multiclass penalty and/or using milestone leveling, the complexity of XP calculation was not the only culprit. Another factor, especially in online play, was that because the SRD included none of the rules for XP or leveling, a lot of people who either didn't own the books or didn't see the point of walking across the room for a book when a website was right there didn't know what they were. In particular, a lot of people were simply unaware that the multiclassing penalty existed at all... and a number of people who did know considered that rule no fun anyway, so ignored it on purpose.

The idea that magic items could be purchased was not just a function of forum culture and price listings, but also of things like the Wealth By Level guidelines and the Community Wealth guidelines, both in the DMG. Notably, the latter explicitly claims that any magic items costing less than a settlement's gp limit can be bought there... though even with my own fondness for Magic-Mart, I wouldn't be quite that free with what can be bought, particularly in smaller settlements. (A village with a 200 gp limit might well have 1st-level potions and oils, but how often will the locals need Oil of Magic Stone instead of Oil of Mending?)

I do like the footnote idea about giving Medium and Heavy loads Arcane Spell Failure chances. I wouldn't make them as severe as actual armor in those ranges, but 10% for a Medium load and 20% for a Heavy load makes sense to me... and if you can ignore ASF for armor of that weight, you can also ignore ASF for a load of that weight.