You might be comparing 2.0's revisions compared to 1.2's. The codebase for 2.0
is much larger than 1.2 was and the chance for "one innocent change" to affect
a lot more subsystems than before is likely to happen. 2.0 is being updated
more actively than 1.2 ever was, so that is also another reason that 2.1 and
2.0 are almost in sync, revision number-wise.
So, I suppose it is all how you look at it.