for my $i ($#$arrayref .. 0) { ... } does not work (initial value of a range must be <= than its terminal value) but

c:\@Work\Perl\monks>perl -wMstrict -le
"my $arrayref = [ qw(uno dos tres quatro cinco) ];
 ;;
 for my $i (reverse 0 .. $#$arrayref) {
   print qq{$i: $arrayref->[$i]};
   }
"
4: cinco
3: quatro
2: tres
1: dos
0: uno
[download]

does and is more readable (again, IMHO).

Having a strange step might be handled by something like

c:\@Work\Perl\monks>perl -wMstrict -le
"my $arrayref = [ qw(uno dos tres quatro cinco) ];
 ;;
 for my $i (grep { $_ % 2 == 0 } 0 .. $#$arrayref) {
   print qq{$i: $arrayref->[$i]};
   }
"
0: uno
2: tres
4: cinco
[download]

which I would tend to favor. The only problem with this kind of loop range expression is that the complete  0 .. $#$arrayref range is built as an intermediate list, so if you're dealing with a large range of any kind you may chew up a lot of memory. The Perl compiler knows enough to optimize a  for my $n (0 .. MAX) { ... } loop to the equivalent C-style for-loop, so no potentially huge list is ever generated; it can't do that trick for  for my $n (grep { whatever } 0 .. MAX) { ... } or similar blocks.

(Another interesting candidate for this sort of thing might be List::MoreUtils::indexes or one of its cousins; also see List::Util.)