The Adobe people knows what makes sense to improve from the standpoint of inside knowledge. Knowing that the image processing pipeline works this way or another, a certain feature might be trivial to implement or near impossible to implement (well). They also have general knowledge from knowing image processing, cpu architectures etc (that they to some degree share with their competitors, but not necessarily with most of their customers).
The Lightroom customers knows what makes sense to them individually. Collectively, they make up a far larger group of Lightroom users than there are developers or beta testers inside Adobe. Thus, one might expect them to be a better estimate of "what annoys current Lightroom users", "what lacking features make them launch external editors" etc. There are some issues with estimating the true collective sentiment (as opposed to only the most vocal forum users).
UI/conceptually, I think that the new HDR/pano feature is somewhat clunky for a non-destructive editor. I would prefer grouping a number of source images (manually and/or automagically based on EXIF), then going directly into "develop" to do tonemapping etc (perhaps add a little grouping-dependent fane in develop for how multiple images are dynamically combined). If (at any point) I chose to add or remove images from the stack, the view in the develop module would simply update. I am sure that there are technical reasons why they do it this way, but hey, I am not paid to write their code :-). Being able to do any mix of HDR/pano/focus stacking in one take would also be sweet.
-h