Approximation: it captures single lines that do not contain @, and multiple lines (if possible) containing patterns that start and end with @.

Not that it helps much, but here it is in action. Anyone have ideas on what it's trying to find?

#! /usr/bin/perl

use strict;
use warnings;

my $file = do {local $/, <DATA>};

my @cvs = $file =~ /((?:(?:[^\n@]+|@[^@]*@)\n?)+)/gs;
print "matched [[$_]]\n" for @cvs;

__DATA__

line 1
line 2
line 3
longer @line@ with @some stuff@
line with nothing on it

line 4
[download]

matched [[line 1
line 2
line 3
longer @line@ with @some stuff@
line with nothing on it
]]
matched [[line 4
]]
[download]