Oh, man.... I'm glad I'm not the only one that came to that conclusion based on Brent's suggestion years ago. Here's a teaser of what I'm working on...
The chart contains the first 5 weeks of what a single index goes through (for 4 different Fill Factors) if it's not based on an "ever increasing" and "immutable" key. Every upswing is a change in page count due to "bad page splits". Every sharp decrease is when the "Best Practice" (do nothing until 10% fragmentation, Reorg between 10 and 30%, Rebuild at 30% or more) kicks in. The Black line is the "Baseline" for an "ever increasing" and "immutable" key, which is 100% only "good page splits". The supposed "Best Practice" method holds back on page splits until the last day or two prior to the over 10% mark and then they all happen at once. This was the cause of huge blocking episodes every Monday (index maintenance occurred every Sunday night) on my main production box.
Guess what happens if you simply stop rebuilding the index?
The Brown Line with the Yellow "glow" is the "No Defrag" line.
I've not finished my performance testing nor have I finished on the exception of rows first being inserted and then expanding the next day or so due to updates but I can tell you that defragging a fragged index that has a 100% Fill Factor is just plain stupid because it causes THOUSANDS of "bad page splits" per day (all of which are blocking episodes and take a heavy toll on the log file) and, remember, this is just for ONE index!!!
The charts for all of this stuff led me to a method to virtually eliminate all pages splits if you actually are addicted to index maintenance. I'm saving that for the presentation I'm giving in May and the article that will follow that.