To answer my questions (according to your links): Yes, normal numbers only have to satisfy conditions on finite strings. And subsequence means deleting some items out of a sequence.

Maybe this was your point, but it seems to me that, with these definition, wtallis's claim (http://news.ycombinator.com/item?id=1567549) is correct; indeed, one simply needs to find 1 at some point, then find 4 a little later, then 1 after that, then 4 again, then a 2, and so forth. The indices where one finds the desired digits become the indices of the claimed subsequence of digits of π.

Indeed. And for the claim to be correct, even a much weaker condition than normality would suffice. E.g. a number like 0.123456789012345678901234567890[repeat] is definitely not normal, but contains all decimal subsequences.