One could argue that it is already in CSV format for a single very big cell ;)

So you could clarify your output as for example 'CSV: one row per hash key or array element, one indentation per substructure'. In which case a simple approach would be, for my own example definition: (Updated to process arrays nested in there)

my $csv = '';
traverse(0,$VAR1,\$csv,';');
sub traverse {
    my ($plevel, $node, $csvref, $dlm) = @_;
    if (ref($node) eq 'HASH')) {
        while (my ($k, $v) = each %$node) {
            $$csvref .= ($plevel x $dlm) . $k;
            if (ref($v)') {
                $$csvref .= "\n";
                traverse($plevel+1, $v, $csvref, $dlm);
            }
            else {
                $$csvref .= $dlm . $v . "\n";
            }
        }
    }
    else { # array. scalars are filtered before the recurse
        for my $elm (@$node) {
            if (ref($elm)) {
                traverse($plevel+1, $elm, $csvref, $dlm);
            }
            else {
                $$csvref .= ($plevel x $dlm) . $elm . "\n";
            }
        }
    }
}
[download]