All the decision making involved with this project have approved in public IRC meetings by the OCT and .LRN technical leadership team, with summaries made available along with the IRC logs.
As far as it being a PITA, we've talked about doing something like this for three? four? more? years because theming OpenACS has always been difficult.
We've delayed doing anything about it because of a lack of resources (people/funding) and the knowledge that in doing so full backwards compatibility would be difficult/impossible to provide.
A big impetus has come from the need to meet accessibility standards. The legal requirement in this regard are strengthening, it needed to be done (well, "needs doing" regarding packages outside .LRN that we want to be "professional").
So ... we decided that, after years of putting it off, the time had come for this rather massive clean-up and rationalization of our software's HTML generation and use of CSS (the fact that we can use CSS and ignore Netscape 4.7 doesn't hurt, either!)
Hopefully the much more sane CSS structure (calendar's custom CSS has gone from 1000+ lines to about 300, for instance, not wonderful but a HUGE improvement) and the noticable ease in theming, plus the infrastructure put in place for accessibility will offset some of the painfulness of the transition people will face when upgrading existing sites.
