Right, I was hurrying to put it up and leave.   Forgot to point out it only works for normal input files where each line ends in a newline.
Actually, your algorithm solves the problem without abandoning regexen.   Unless we're missing something, Larry was thinking along the lines of bart's logic above, that you have to know the new length of the line up to this point.   But you are just depending on the text in front of the present tab(s) to be aligned either with the beginning of the line or the previous tab stop, so there is no need to consider the entire preceding line segment:
$ perl -we'$_="\t\tYou\tare\tmaking\t\ttoo\tmany\tassumptions...\t\t";
s/(^|[^\t]+)(\t+)/$1." " x (length($2) * 8 - (length($1) & 7))/ge;
You are making too many assump