Hi all. Rich sent me a PM asking for an additional summary vs. what I posted near the top of this thread, so this is what I sent him, and thought I should repost the essentials:
I hope I did a good job outlining the complete situation in the original blog post: http://www.sencha.com/blog/ext-js-4-1-update/ and my follow-up forum posts and comments. But to summarize again, if 4.x performance is not meeting your needs then for the short term you should stay on the 3.4x line. We're looking at extending 3.x support for at least 6 months beyond the 5.x ship based on this recommendation, since more people will be staying on 3.x for longer.
Our goal is to get the 4.x line to be faster than 3.4 at the very least for the common app cases. From looking at what people are posting on this thread and our internal testing, it looks like the consensus is that the current 4.1 preview is roughly 2x faster than 4.0 but also still about 2x slower than 3.4. We are working as aggressively as possible and will continue to release performance improvements in 4.1.x and 4.2 releases until we achieve our performance goals. Hence our recommendation. As an aside, we've integrated continuous performance benchmarking in our testing process as a result of our 4.0 performance experience.
Our initial comments that 4.x would be faster than 3.x was based on early benchmarks of our refactored layout and rendering pipeline - which we showed at SenchaCON 2010. Subsequent to SenchaCON 201, we integrated the new class system, added fairly involved theming code and also did a grid rewrite to support the new 4.x features. The combination of these, particularly for more complex layouts, reduced performance substantially from our early benchmarks, as all of our, and your, test benchmarks have shown.
I hope this summary helps those of you who are trying to make decisions about which version of Ext JS to use.