What about good ol' line by line processing?

use strict;
use warnings;
use Data::Dumper;

my $t = q{
Subject: Hi there
Content-Type: text/plain;
    format=flowed;
    charset="iso-1189-1";
Return-Path: <example@example.org>
};

my @fields;
for ( split /\n/, $t ) {
  push @fields, $_ and next if /^\S+:/;
  $fields[-1] .= $1 if /\s*(.+)/;
}
print Dumper \@fields;
[download]