Haxe did nothing you couldn't do with AS3 and Alchemy.
>With all of these issues known to Adobe you might suspect that they would move aggressively to increase the performance of Flash in general and Flex in particular
They did, by quite a bit. They also had parallel project for increasing performance with Stage3D and Alchemy.
I get your point about Flex 4. It is a better UI framework than Flex 3, but not significantly better, so the engineering effort that went toward it could have been better spent doing something else.
I also mirror your thoughts about "what could have been". My pet theory is that Stage3D should have been a FP10 feature (which would put it around the beginning of 2009 timeframe) and Adobe should have had some sort of (optional) browser-based app store to push Flash as a gaming platform. And since we're in hypothetical scenarios, AS3 should have been better designed, and cross-compilation to HTML/JS should have been there from the beginning. This may have given Flash some legs. But the reality is that sooner or later it would have been superseded by HTML anyway. Oh well, hindsight.
Furthermore, looking at the performance gains that Google, Microsoft and Mozilla managed to squeeze out of a language like JS, one can't help but think how much faster ABC and Flash Runtime could been.