Crossposted question deserves a crossposted answer. Here's what I came up with:

#!/usr/bin/env perl
use strict;
use warnings;
use Data::Dumper;
use 5.14.0;

my %stuff;

#extract the header row.
#use the regex to remove the linefeed, because
#we can't chomp it inline like this. 
#works since perl 5.14
#otherwise we could just chomp (@header) later. 
my ( $id,  @header ) = split( /\|/, <DATA> );
chomp (@header);

while (<DATA>) {
    #turn this row into a hash of key-values.
    my %row;
    ( $id, @row{@header} ) = split(/\|/);
    #diag print
    print Dumper \%row;

    #iterate each key, and insert into $row.
    foreach my $key ( keys %row ) {
        push( @{ $stuff{$id}{$key} }, $row{$key} );
    }
}

#diag print
print Dumper \%stuff;

foreach my $id ( sort keys %stuff ) {

    #join this record by '|'.
    print join('|',
        $id,
        #turn inner arrays into comma separated via map.
        map {
            my %seen;
            #use grep to remove dupes - e.g. "abc,abc" -> "abc"
            join( ",", grep !$seen{$_}++, @$_ )
        } @{ $stuff{$id} }{@header}
        ),
        "\n";
}

__DATA__
id|Name|app1|app2
1|abc|234|231|
2|xyz|123|215|
1|abc|265|321|
3|asd|213|235|
[download]

• We iterate each line of the input, and make a hash of arrays, keyed on `id`.
• Then we iterate the hash of arrays, and use map/join/grep to remove duplicates, collapse the arrays into strings (So we don't end up with an "abc,abc" field)

As a result it'll generate:

id|Name|app1|app2
1|abc|234,265|231,321
2|xyz|123|215
3|asd|213|235
[download]

Therefore handles arbitrary fields and arbitrary numbers of 'collisions' neatly.