I took Mary's comment to mean moving the themes that come packed with Moodle over to inheriting from Bootstrap rather than Base. Which I think is a very achievable goal for 2.6 and, since they're themes and so disconnected from core, could even be done before then and used by the more adventurous types shortly after 2.5 goes live. Indeed it would probably be best if they worked in 2.3 and 2.4 as well to get the long-term migration started as early as possible.
Quite a few people are asking about the best way to extend Bootstrap in Moodle so writing a how-to showing the benefits of the new approach is very high on my todo list. It'll be loosely based on this article, by the chap behind Bootswatch.com: http://coding.smashingmagazine.com/2013/03/12/customizing-bootstrap/ but with an example starter Moodle theme for people to build on as well.
Porting some of the themes that come with Moodle (as well as some from the Bootstrap community) as a demo would probably be a part of that, and I'd be hopeful that current Moodle Themers would join in and try porting their own themes, as they'll probably provide the most informed feedback on what needs to be improved. It's also the same technique I'll be using for our institution's theme when updating this summer to 2.4 so I'm personally invested in it working well.
But yes, you can't actually "lose" Base & Canvas until everyone that builds on them moves on. On the other hand, you may want to deprecate them and stop putting full effort into them, possibly fairly soon. And I think that's perhaps what Mary meant.
There's been a few conversations recently where people have said things like "get rid of", "kill" etc. when they really meant "decide to build on something else for the future and deprecate this in an orderly manner without unduly affecting those who rely on it while transferring as many people as quickly as possible to take advantage of the new way to our/their collective benefit". It's just not as much fun to say it that way but it is very important to communicate that well to avoid unnecessarily worrying people.
As for "Bootstrap" vs "Any modern HTML/CSS framework", they overlap so much that I think it is fair to consider them pretty much interchangeable in most conversations. Doubly so if you consider that Bootstrap is transitioning from 2.3 to 3 (and the latter version should probably be targeted for 2.6) so even just addressing "Bootstrap" alone you can paint yourself into a corner if you take too many shortcuts (though some people seem a bit freaked out by the thought of Bootstrap changing version and I don't mean to add to that, it's really not that dramatic an upgrade, more a streamlining of what's currently there). Bootstrap does however present a very concrete rallying point and reference for collaboration, much more so than a vaguer, more abstract goal would.
The key thing, imho, is to have more renderers at the right level of granularity and I think working towards a goal of implementing Bootstrap (which is fundamentally a collection of re-usable components) is the best way to get there quickly.