RE: "non-zero top border" in 8.3.1

I think I (almost) understand now the logic of "top border" rule. In some cases it would put elements higher than where they would be if they had content or bottom margin, which would compensate for their bottom margin now collapsing with their top margin, and leave more space for any nested floats.

I am not convinced yet that I fully understand it yet though. I see 3 problems here:

1) without the "top border" rule, there is some continuity in position of empty elements (see attached sequence): removing element's content normally doesn't change its position. It may shift it vertically when its bottom border collapses with its top border

2) when "top border" rule is added, it has to be understood when exactly it happens. Does the previous rule ("If the element's margins are collapsed with its parent's top margin, the top border edge of the box is defined to be the same as the parent's.") still apply? But that's not what would happen if the parent had a top border. But if it didn't apply, expected behavior in test 68 would not be correct.

3) I don't really have a preference towards any existing implementation (including that of IE7). I want to find what is considered correct by the standard (existing interop is a bonus but not really required). I am not convinced that current browsers have the same position though. *DBarron*, is it fine with you to say that Mozilla's is incorrect in
http://www.hixie.ch/tests/adhoc/css/box/block/margin-collapse/068.html
?