The push for 25nm by IMFT was based on lower costs and higher capacity. Both Intel and Micron have stated this publically and it was the sole objective of the investment in 25nm technology.

NAND is a multi billion industry. Granted most of that NAND does not end up in SSD's, but if there are not economies in scale now when will there be?

If you are developing features on-die it is unlikely that you would embed them without fully engineering them first, especially if your road map planned to use the features in future releases. I'm not knowledgeable about such things, but it would seem unlikely that you would embedded a technology unless you knew it would work.