Approaches:

1.Metaprogramming

Generate dynamically a string of the above code with all nesting and eval it. (fast and intuitive!)

2. Recursion

Write a function X2() which gets two arrayrefs and returns an arr_ref of the cross product. Call it recursively until @array exceeded.

3. Reduce

is a variation of the above, with List::Util::reduce

use Data::Dumper;
 
#- crossproduct of two array refs
sub X2 {
  my ($a,$b)=@_;
  my @result;
  for my $x (@$a) {
    for my $y (@$b) {
      unless (ref($x) eq "ARRAY"){
        push @result, [$x,$y];
      }else{
        push @result, [(@$x,$y)];
      }      
    }
  }
  return \@result;
}
 
use List::Util qw/reduce/;
 
#- crossproduct of list of array refs
sub X { reduce { X2($a,$b) } @_ }
 
my @array = (
    [ "a", "b", "c", ],
    [ "1", "2", "3", "4", ],
    [ "x", "y", ],
); 
 
print Dumper X(@array);
[download]

The function X() now works somehow like the X operator in perl6, but of course you could also use reduce { X2($a,$b) } @array directly.

The function X2() could be rewritten with two nested maps, but thats a little two cryptic for my taste.

Bad ideas:

1. Glob

that's a hack which only works for strings as element type, everything else will be stringified, eg refs!!!

UPDATES:

* just noticed that you only want a simple concatenation of strings. That simplifies the code...

* There is a problem with this code .. the first one to spot it gets upvoted! :)