I'm not arguing that it was in some way a failure; that is a matter of public record. 3.0 was the failure, given its design was screwed-up enough that the only way they could fix it was to release a half-edition.
Most of the ongoing success since then isn't a credit to WotC. WotC was actually a primary source of a lot of 3E's problems; they had poor quality control measures, were constantly fighting with 3PP companies (even going as far as to revise the OGL to try to stop some 3PPs from putting out products), had power creep that still makes people cringe, stopped caring about even producing quality errata and at least twice had to errata their errata before that, and had a lot of other problems.
Pretty much, Paizo had to work with a broken 3E fanbase that was fighting itself and try to forge a product out of an epic mess that was left behind when WotC decided to move on to 4E. At that point, it was no longer in WotC's hands and WotC could not claim any credit for what Paizo did.
So, for WotC, what I said is true; it was one of their editions with a pretty massive failure in it. For Paizo, what I said doesn't apply, since they got it right on their first try.
I tend to agree with your evidence, but still wouldn't say that 3.X was a failure. It was a flawed success. A success at a huge price....
For those who don't see it, the symptoms of 3.0E as a flawed success:
- The need for a half-edition
- The rise of the OSR and its resultant loss of customers
- The rise of electronic publication by 3pp
- The more radical surgery by Paizo to make 3.5 succeed
- The Rise of 3PP on the backs of the OGL to produce new non-d20 games in competition with the D&D brand
- The collapse of the 3pp market
The OSR stripped a bunch of loyal and a large number of potential customers. Brand loyalty overridden.
3pp going electronic: bad for the LGS... and at the heart of it, Paizo considered the end customer to be their customer, while WotC did and still does consider the LGS their customers, with the end customer being a good co-sold with the games to the LGS. Bad for the LGS is thus bad for WotC.
Pathfinder: It's not just the 3.5 SRD. It's a significantly modified 3.5 SRD. And that means that Pathfinder's success isn't just "We kept 3.X in print and succeeded." It was, in fact, "We created a new game upon the 3.x skeleton, and fixed the majority of the problems, while building a new brand loyalty."
3pp publishers going non-d20: Mongoose, Green Ronin, and others have become competitors, rather than supporters, of the D&D brand. Mongoose Traveller, Mongoose RuneQuest; Green Ronin's Dragon Age RPG, even MCG's Numenara...
The collapse of the 3pp market a decade ago was also firmly a function of failure by Wizards. After opening up the market to 3pp, 3.x became associated in retail with the stable selling corebooks, and the piles of worthless drekh from the 3PP... It's a symptom of a bad game design that the various uses it was put to were all too often incompatible with each other.... Many LGSs lost big due to the unmovable stock. This collapse wasn't entirely Wizards' fault, but was an unintended consequence of the OGL. Too many produced too much too fast, and many produced stuff that wasn't worthwhile. It was hard to tell the good from the bad. And so a lot of the good didn't get onto shelves, either. And a lot of what did make it didn't sell. Why? No controls on the 3pp market.
Let's compare to the videogame industry: The PS, X-box, and Wii markets are largely 3PP... but there is a quality control in place. So, for the most part, the standard packaging means a game that's at least relatively stable, relatively playable, and runs on the listed hardware. When Atari didn't do this, it killed brand loyalty. That was in the early 1980's. Atari never recovered, despite really good hardware that came out following the collapse of the 2600... Nintendo, in the late 80's, knew that 3PP would propel their platform, but they were smart enough to require an approval process before indications of compatibility using their logos...
So, WotC had a model to see what might happen with the open license... but didn't pay attention. And not one, but two different but closely related companies.